In [ ]:
fig, ax = plt.subplots(figsize=(10, 7), subplot_kw=dict(projection='3d'))
ax.plot_surface(x1, x2, y)
ax.set(
xlabel='$x_1$',
ylabel='$x_2$',
zlabel='$f(x_1, x_2)$'
)
plt.tight_layout()
plt.show()
import numpy as np
import matplotlib.pyplot as plt
import random
import pandas as pd
import tensorflow as tf
from sklearn.model_selection import train_test_split
from sklearn.neural_network import MLPRegressor
from sklearn.metrics import mean_squared_error
from sklearn.metrics import mean_absolute_error
from sklearn.preprocessing import StandardScaler
from sklearn.preprocessing import Normalizer
from tensorflow.keras.layers import Dense
from tensorflow.keras.models import Sequential
from tensorflow.keras.optimizers import SGD, Adam
from keras.layers import Flatten,Dense,BatchNormalization,Activation,Dropout
from keras.callbacks import ReduceLROnPlateau
from sklearn.metrics import classification_report, confusion_matrix, ConfusionMatrixDisplay
Utilize redes neurais perceptron de múltiplas camadas para aproximar as funções abaixo. Apresente um gráfico com a curva da função analítica e a curva da função aproximada pela rede neural. Apresente também a curva do erro médio de treinamento com relação ao número de épocas e a curva do erro médio com o conjunto de validação. Procure definir para cada função a arquitetura da rede neural perceptron, isto é, o número de entradas, o número de neurônios em cada camada e o número de neurônios camada de saída.
Observações: Como se trata de um problema de aproximação de funções, considere a ca- mada de saída do tipo linear puro. Isto é, φ(v)=v, onde v é o potencial de ativação.
$f(x_1,x_2) = (1 -x_1)^2 + 100 (x_2 - (x_1)^2)^2$ com $-10 ≤ x_1 ≤ 10, -10 ≤ x_2 ≤10$
# function's definition
def f(x1,x2):
return (1 - x1)**2 + 100*(x2 - (x1)**2)**2
# generating points
x1, x2 = np.meshgrid(np.linspace(-10, 10, 100), np.linspace(-10, 10, 100))
y = f(x1,x2)
fig, ax = plt.subplots(figsize=(10, 7), subplot_kw=dict(projection='3d'))
ax.plot_surface(x1, x2, y)
ax.set(
xlabel='$x_1$',
ylabel='$x_2$',
zlabel='$f(x_1, x_2)$'
)
plt.tight_layout()
plt.show()
# split training and test
x_train, x_test, y_train, y_test = train_test_split(
np.vstack([x1.flatten(), x2.flatten()]).T,
y.flatten(),
test_size=0.2,
random_state=42
)
fig, ax = plt.subplots(figsize=(10, 7), subplot_kw=dict(projection='3d'))
ax.plot_wireframe(x1, x2, y, linewidths=0.5, color='lightgrey')
ax.scatter(x_train[:,0], x_train[:,1], y_train, s=1, color='darkorange', label='Training data')
ax.scatter(x_test[:,0], x_test[:,1], y_test, s=5, color='darkgreen', label='Test data')
ax.set(
xlabel='$x_1$',
ylabel='$x_2$',
zlabel='$f(x_1, x_2)$'
)
plt.legend()
plt.tight_layout()
plt.show()
# created scaler
scaler = StandardScaler()
y_train = y_train.reshape(-1,1)
y_test = y_test.reshape(-1,1)
# fit normalizer on training dataset
scaler.fit(y_train)
# transform training dataset
y_train = scaler.transform(y_train)
# transform test dataset
y_test = scaler.transform(y_test)
# verifying shapes
print("X train shape: ", x_train.shape)
print("Y train shape: ", y_train.shape)
print("X test shape: ", x_test.shape)
print("Y test shape: ", y_test.shape)
X train shape: (8000, 2) Y train shape: (8000, 1) X test shape: (2000, 2) Y test shape: (2000, 1)
mlp = Sequential([
Dense(64, activation='relu', input_shape=(2,)),
Dense(32, activation='relu'),
Dense(16, activation='relu'),
Dense(8, activation='relu'),
Dense(4, activation='relu'),
Dense(1, activation='linear')
])
mlp.compile(
loss='mean_squared_error',
optimizer='adam'
)
mlp.summary()
Model: "sequential_1"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
dense_6 (Dense) (None, 64) 192
dense_7 (Dense) (None, 32) 2080
dense_8 (Dense) (None, 16) 528
dense_9 (Dense) (None, 8) 136
dense_10 (Dense) (None, 4) 36
dense_11 (Dense) (None, 1) 5
=================================================================
Total params: 2,977
Trainable params: 2,977
Non-trainable params: 0
_________________________________________________________________
history = mlp.fit(
x_train, y_train,
batch_size=8,
epochs=2000,
validation_split=0.1,
callbacks=[
tf.keras.callbacks.EarlyStopping(monitor='val_loss', patience=10),
tf.keras.callbacks.ReduceLROnPlateau(monitor='val_loss', factor=0.1, patience=5, min_lr=0.0001)
]
)
Epoch 1/2000 900/900 [==============================] - 2s 1ms/step - loss: 0.2403 - val_loss: 0.0481 - lr: 0.0010 Epoch 2/2000 900/900 [==============================] - 1s 1ms/step - loss: 0.0301 - val_loss: 0.0225 - lr: 0.0010 Epoch 3/2000 900/900 [==============================] - 1s 1ms/step - loss: 0.0200 - val_loss: 0.0144 - lr: 0.0010 Epoch 4/2000 900/900 [==============================] - 1s 1ms/step - loss: 0.0144 - val_loss: 0.0141 - lr: 0.0010 Epoch 5/2000 900/900 [==============================] - 1s 1ms/step - loss: 0.0098 - val_loss: 0.0073 - lr: 0.0010 Epoch 6/2000 900/900 [==============================] - 1s 1ms/step - loss: 0.0094 - val_loss: 0.0056 - lr: 0.0010 Epoch 7/2000 900/900 [==============================] - 1s 1ms/step - loss: 0.0102 - val_loss: 0.0024 - lr: 0.0010 Epoch 8/2000 900/900 [==============================] - 1s 1ms/step - loss: 0.0082 - val_loss: 0.0071 - lr: 0.0010 Epoch 9/2000 900/900 [==============================] - 1s 1ms/step - loss: 0.0055 - val_loss: 0.0031 - lr: 0.0010 Epoch 10/2000 900/900 [==============================] - 1s 1ms/step - loss: 0.0053 - val_loss: 0.0126 - lr: 0.0010 Epoch 11/2000 900/900 [==============================] - 1s 1ms/step - loss: 0.0051 - val_loss: 8.4492e-04 - lr: 0.0010 Epoch 12/2000 900/900 [==============================] - 1s 1ms/step - loss: 0.0045 - val_loss: 0.0091 - lr: 0.0010 Epoch 13/2000 900/900 [==============================] - 1s 1ms/step - loss: 0.0103 - val_loss: 7.3677e-04 - lr: 0.0010 Epoch 14/2000 900/900 [==============================] - 1s 1ms/step - loss: 0.0041 - val_loss: 0.0061 - lr: 0.0010 Epoch 15/2000 900/900 [==============================] - 1s 1ms/step - loss: 0.0034 - val_loss: 0.0015 - lr: 0.0010 Epoch 16/2000 900/900 [==============================] - 1s 1ms/step - loss: 0.0047 - val_loss: 0.0060 - lr: 0.0010 Epoch 17/2000 900/900 [==============================] - 1s 1ms/step - loss: 0.0028 - val_loss: 0.0046 - lr: 0.0010 Epoch 18/2000 900/900 [==============================] - 1s 1ms/step - loss: 0.0045 - val_loss: 0.0064 - lr: 0.0010 Epoch 19/2000 900/900 [==============================] - 1s 1ms/step - loss: 3.8528e-04 - val_loss: 3.2380e-04 - lr: 1.0000e-04 Epoch 20/2000 900/900 [==============================] - 1s 1ms/step - loss: 2.4806e-04 - val_loss: 2.2849e-04 - lr: 1.0000e-04 Epoch 21/2000 900/900 [==============================] - 1s 1ms/step - loss: 2.4695e-04 - val_loss: 2.0553e-04 - lr: 1.0000e-04 Epoch 22/2000 900/900 [==============================] - 1s 1ms/step - loss: 2.3648e-04 - val_loss: 1.9860e-04 - lr: 1.0000e-04 Epoch 23/2000 900/900 [==============================] - 1s 1ms/step - loss: 2.3406e-04 - val_loss: 1.8407e-04 - lr: 1.0000e-04 Epoch 24/2000 900/900 [==============================] - 1s 1ms/step - loss: 2.1895e-04 - val_loss: 2.1349e-04 - lr: 1.0000e-04 Epoch 25/2000 900/900 [==============================] - 1s 1ms/step - loss: 2.5439e-04 - val_loss: 2.7080e-04 - lr: 1.0000e-04 Epoch 26/2000 900/900 [==============================] - 1s 1ms/step - loss: 2.3294e-04 - val_loss: 3.1382e-04 - lr: 1.0000e-04 Epoch 27/2000 900/900 [==============================] - 1s 1ms/step - loss: 1.9288e-04 - val_loss: 2.5049e-04 - lr: 1.0000e-04 Epoch 28/2000 900/900 [==============================] - 2s 2ms/step - loss: 2.0941e-04 - val_loss: 1.4517e-04 - lr: 1.0000e-04 Epoch 29/2000 900/900 [==============================] - 2s 2ms/step - loss: 2.0231e-04 - val_loss: 1.7252e-04 - lr: 1.0000e-04 Epoch 30/2000 900/900 [==============================] - 1s 1ms/step - loss: 1.8211e-04 - val_loss: 1.2884e-04 - lr: 1.0000e-04 Epoch 31/2000 900/900 [==============================] - 1s 1ms/step - loss: 1.8255e-04 - val_loss: 1.1797e-04 - lr: 1.0000e-04 Epoch 32/2000 900/900 [==============================] - 1s 1ms/step - loss: 1.5825e-04 - val_loss: 2.3869e-04 - lr: 1.0000e-04 Epoch 33/2000 900/900 [==============================] - 1s 1ms/step - loss: 1.8320e-04 - val_loss: 1.7549e-04 - lr: 1.0000e-04 Epoch 34/2000 900/900 [==============================] - 1s 1ms/step - loss: 1.5552e-04 - val_loss: 1.2200e-04 - lr: 1.0000e-04 Epoch 35/2000 900/900 [==============================] - 1s 1ms/step - loss: 1.6122e-04 - val_loss: 1.1288e-04 - lr: 1.0000e-04 Epoch 36/2000 900/900 [==============================] - 1s 1ms/step - loss: 1.6115e-04 - val_loss: 1.4661e-04 - lr: 1.0000e-04 Epoch 37/2000 900/900 [==============================] - 1s 1ms/step - loss: 1.7000e-04 - val_loss: 1.8074e-04 - lr: 1.0000e-04 Epoch 38/2000 900/900 [==============================] - 1s 1ms/step - loss: 1.5897e-04 - val_loss: 1.3702e-04 - lr: 1.0000e-04 Epoch 39/2000 900/900 [==============================] - 2s 2ms/step - loss: 1.8850e-04 - val_loss: 9.9578e-05 - lr: 1.0000e-04 Epoch 40/2000 900/900 [==============================] - 1s 1ms/step - loss: 1.8244e-04 - val_loss: 1.0522e-04 - lr: 1.0000e-04 Epoch 41/2000 900/900 [==============================] - 1s 1ms/step - loss: 1.3757e-04 - val_loss: 1.5331e-04 - lr: 1.0000e-04 Epoch 42/2000 900/900 [==============================] - 1s 1ms/step - loss: 1.3224e-04 - val_loss: 2.4568e-04 - lr: 1.0000e-04 Epoch 43/2000 900/900 [==============================] - 1s 1ms/step - loss: 1.6857e-04 - val_loss: 1.4375e-04 - lr: 1.0000e-04 Epoch 44/2000 900/900 [==============================] - 1s 1ms/step - loss: 1.3927e-04 - val_loss: 3.8073e-04 - lr: 1.0000e-04 Epoch 45/2000 900/900 [==============================] - 1s 1ms/step - loss: 1.6187e-04 - val_loss: 4.2486e-04 - lr: 1.0000e-04 Epoch 46/2000 900/900 [==============================] - 1s 1ms/step - loss: 1.4784e-04 - val_loss: 1.7491e-04 - lr: 1.0000e-04 Epoch 47/2000 900/900 [==============================] - 1s 1ms/step - loss: 1.5885e-04 - val_loss: 1.2637e-04 - lr: 1.0000e-04 Epoch 48/2000 900/900 [==============================] - 1s 1ms/step - loss: 1.3876e-04 - val_loss: 9.0454e-05 - lr: 1.0000e-04 Epoch 49/2000 900/900 [==============================] - 1s 1ms/step - loss: 1.7796e-04 - val_loss: 1.3724e-04 - lr: 1.0000e-04 Epoch 50/2000 900/900 [==============================] - 2s 2ms/step - loss: 1.3802e-04 - val_loss: 1.1495e-04 - lr: 1.0000e-04 Epoch 51/2000 900/900 [==============================] - 2s 2ms/step - loss: 1.4784e-04 - val_loss: 1.0661e-04 - lr: 1.0000e-04 Epoch 52/2000 900/900 [==============================] - 1s 1ms/step - loss: 1.2565e-04 - val_loss: 9.0812e-05 - lr: 1.0000e-04 Epoch 53/2000 900/900 [==============================] - 1s 1ms/step - loss: 1.4147e-04 - val_loss: 1.0544e-04 - lr: 1.0000e-04 Epoch 54/2000 900/900 [==============================] - 1s 1ms/step - loss: 1.4119e-04 - val_loss: 5.1870e-04 - lr: 1.0000e-04 Epoch 55/2000 900/900 [==============================] - 1s 1ms/step - loss: 1.4186e-04 - val_loss: 7.9401e-05 - lr: 1.0000e-04 Epoch 56/2000 900/900 [==============================] - 1s 1ms/step - loss: 1.7324e-04 - val_loss: 9.3971e-05 - lr: 1.0000e-04 Epoch 57/2000 900/900 [==============================] - 1s 1ms/step - loss: 1.3775e-04 - val_loss: 1.4554e-04 - lr: 1.0000e-04 Epoch 58/2000 900/900 [==============================] - 1s 1ms/step - loss: 1.2992e-04 - val_loss: 9.5643e-05 - lr: 1.0000e-04 Epoch 59/2000 900/900 [==============================] - 1s 1ms/step - loss: 1.2934e-04 - val_loss: 7.1712e-05 - lr: 1.0000e-04 Epoch 60/2000 900/900 [==============================] - 1s 1ms/step - loss: 1.3317e-04 - val_loss: 1.5182e-04 - lr: 1.0000e-04 Epoch 61/2000 900/900 [==============================] - 1s 1ms/step - loss: 1.2192e-04 - val_loss: 9.4130e-05 - lr: 1.0000e-04 Epoch 62/2000 900/900 [==============================] - 1s 1ms/step - loss: 1.2367e-04 - val_loss: 1.1338e-04 - lr: 1.0000e-04 Epoch 63/2000 900/900 [==============================] - 1s 1ms/step - loss: 1.5802e-04 - val_loss: 1.1068e-04 - lr: 1.0000e-04 Epoch 64/2000 900/900 [==============================] - 1s 1ms/step - loss: 1.1099e-04 - val_loss: 1.8395e-04 - lr: 1.0000e-04 Epoch 65/2000 900/900 [==============================] - 1s 1ms/step - loss: 1.2799e-04 - val_loss: 6.3500e-05 - lr: 1.0000e-04 Epoch 66/2000 900/900 [==============================] - 1s 1ms/step - loss: 1.2981e-04 - val_loss: 1.4402e-04 - lr: 1.0000e-04 Epoch 67/2000 900/900 [==============================] - 1s 1ms/step - loss: 1.2969e-04 - val_loss: 6.7762e-05 - lr: 1.0000e-04 Epoch 68/2000 900/900 [==============================] - 1s 1ms/step - loss: 1.3877e-04 - val_loss: 9.6731e-05 - lr: 1.0000e-04 Epoch 69/2000 900/900 [==============================] - 1s 1ms/step - loss: 1.1277e-04 - val_loss: 1.3771e-04 - lr: 1.0000e-04 Epoch 70/2000 900/900 [==============================] - 1s 1ms/step - loss: 1.1472e-04 - val_loss: 1.2268e-04 - lr: 1.0000e-04 Epoch 71/2000 900/900 [==============================] - 1s 1ms/step - loss: 1.1190e-04 - val_loss: 7.2661e-05 - lr: 1.0000e-04 Epoch 72/2000 900/900 [==============================] - 1s 1ms/step - loss: 1.3607e-04 - val_loss: 1.3418e-04 - lr: 1.0000e-04 Epoch 73/2000 900/900 [==============================] - 1s 1ms/step - loss: 1.1247e-04 - val_loss: 2.3557e-04 - lr: 1.0000e-04 Epoch 74/2000 900/900 [==============================] - 1s 1ms/step - loss: 1.1410e-04 - val_loss: 9.0923e-05 - lr: 1.0000e-04 Epoch 75/2000 900/900 [==============================] - 1s 1ms/step - loss: 1.1888e-04 - val_loss: 2.2146e-04 - lr: 1.0000e-04
fig, ax = plt.subplots(figsize=(10, 7))
ax.plot(history.history['loss'], label='Training Loss')
ax.plot(history.history['val_loss'], label='Validation Loss')
ax.set(
title='Processo de otimização das funções de perda',
ylabel='Loss',
xlabel='Epoch'
)
plt.legend()
plt.tight_layout()
plt.show()
predictions = mlp.predict(x_test)
inverse_predictions = scaler.inverse_transform(predictions)
63/63 [==============================] - 0s 996us/step
fig, ax = plt.subplots(figsize=(15, 9), subplot_kw=dict(projection='3d'))
ax.plot_wireframe(x1, x2, y, linewidths=0.5, color='lightgrey')
ax.scatter(x_test[:,0], x_test[:,1], scaler.inverse_transform(y_test), s=14, color='C0', label='Training data')
ax.scatter(x_test[:,0], x_test[:,1], inverse_predictions, s=15, marker='^', color='C1', label='Test data')
ax.set(
xlabel='$x_1$',
ylabel='$x_2$',
zlabel='$f(x_1, x_2)$'
)
plt.legend()
plt.tight_layout()
plt.show()
mse = mean_squared_error(scaler.inverse_transform(y_test), inverse_predictions)
rmse = mean_squared_error(scaler.inverse_transform(y_test), inverse_predictions, squared=False)
mae = mean_absolute_error(scaler.inverse_transform(y_test), inverse_predictions)
print(f"Mean Squared Error: {mse}")
print(f"Root Mean Squared Error: {rmse}")
print(f"Mean Absolute Error: {mae}")
Mean Squared Error: 16880097.554985106 Root Mean Squared Error: 4108.5395890736045 Mean Absolute Error: 2477.902080862183
$x_1^2 + x_2^2 + 2x_1x_2cos(\pi x_1x_2) + x_1 + x_2 -1$ com $|x_1| ≤1 , |x_2| ≤ 1$
# function's definition
def g(x1,x2):
return x1**2 + x2**2 + 2 * x1 * x2 * np.cos(np.pi * x1 * x2) + x1 + x2 - 1
# generating points
x1, x2 = np.meshgrid(np.linspace(-1, 1, 100), np.linspace(-1, 1, 100))
y = g(x1,x2)
# plotting surface
fig, ax = plt.subplots(figsize=(10, 7), subplot_kw=dict(projection='3d'))
ax.plot_surface(x1, x2, y)
ax.set(
xlabel='$x_1$',
ylabel='$x_2$',
zlabel='$f(x_1, x_2)$'
)
plt.tight_layout()
plt.show()
# split train test
x_train, x_test, y_train, y_test = train_test_split(
np.vstack([x1.flatten(), x2.flatten()]).T,
y.flatten(),
test_size=0.2,
random_state=505
)
# surface os points to test and to train
fig, ax = plt.subplots(figsize=(10, 7), subplot_kw=dict(projection='3d'))
ax.plot_wireframe(x1, x2, y, linewidths=0.5, color='lightgrey')
ax.scatter(x_train[:,0], x_train[:,1], y_train, s=1, color='darkorange', label='Training data')
ax.scatter(x_test[:,0], x_test[:,1], y_test, s=5, color='darkgreen', label='Test data')
ax.set(
xlabel='$x_1$',
ylabel='$x_2$',
zlabel='$f(x_1, x_2)$'
)
plt.legend()
plt.tight_layout()
plt.show()
mlp = Sequential([
Dense(64, activation='relu', input_shape=(2,)),
Dense(32, activation='relu'),
Dense(16, activation='relu'),
Dense(8, activation='relu'),
Dense(4, activation='relu'),
Dense(1, activation='linear')
])
mlp.compile(
loss='mean_squared_error',
optimizer='adam'
)
mlp.summary()
Model: "sequential_3"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
dense_18 (Dense) (None, 64) 192
dense_19 (Dense) (None, 32) 2080
dense_20 (Dense) (None, 16) 528
dense_21 (Dense) (None, 8) 136
dense_22 (Dense) (None, 4) 36
dense_23 (Dense) (None, 1) 5
=================================================================
Total params: 2,977
Trainable params: 2,977
Non-trainable params: 0
_________________________________________________________________
# training the model
history = mlp.fit(
x_train, y_train,
batch_size=8,
epochs=2000,
validation_split=0.1,
callbacks=[
tf.keras.callbacks.EarlyStopping(monitor='val_loss', patience=10),
tf.keras.callbacks.ReduceLROnPlateau(monitor='val_loss', factor=0.1, patience=5, min_lr=0.0001)
]
)
Epoch 1/2000 900/900 [==============================] - 2s 1ms/step - loss: 0.3755 - val_loss: 0.1831 - lr: 0.0010 Epoch 2/2000 900/900 [==============================] - 1s 1ms/step - loss: 0.1221 - val_loss: 0.0754 - lr: 0.0010 Epoch 3/2000 900/900 [==============================] - 1s 1ms/step - loss: 0.0488 - val_loss: 0.0392 - lr: 0.0010 Epoch 4/2000 900/900 [==============================] - 1s 1ms/step - loss: 0.0238 - val_loss: 0.0227 - lr: 0.0010 Epoch 5/2000 900/900 [==============================] - 1s 1ms/step - loss: 0.0188 - val_loss: 0.0208 - lr: 0.0010 Epoch 6/2000 900/900 [==============================] - 1s 1ms/step - loss: 0.0147 - val_loss: 0.0175 - lr: 0.0010 Epoch 7/2000 900/900 [==============================] - 1s 1ms/step - loss: 0.0147 - val_loss: 0.0202 - lr: 0.0010 Epoch 8/2000 900/900 [==============================] - 1s 1ms/step - loss: 0.0169 - val_loss: 0.0165 - lr: 0.0010 Epoch 9/2000 900/900 [==============================] - 1s 1ms/step - loss: 0.0135 - val_loss: 0.0165 - lr: 0.0010 Epoch 10/2000 900/900 [==============================] - 1s 1ms/step - loss: 0.0148 - val_loss: 0.0140 - lr: 0.0010 Epoch 11/2000 900/900 [==============================] - 1s 1ms/step - loss: 0.0100 - val_loss: 0.0106 - lr: 0.0010 Epoch 12/2000 900/900 [==============================] - 1s 1ms/step - loss: 0.0096 - val_loss: 0.0085 - lr: 0.0010 Epoch 13/2000 900/900 [==============================] - 1s 1ms/step - loss: 0.0058 - val_loss: 0.0064 - lr: 0.0010 Epoch 14/2000 900/900 [==============================] - 1s 1ms/step - loss: 0.0059 - val_loss: 0.0047 - lr: 0.0010 Epoch 15/2000 900/900 [==============================] - 1s 1ms/step - loss: 0.0049 - val_loss: 0.0043 - lr: 0.0010 Epoch 16/2000 900/900 [==============================] - 1s 1ms/step - loss: 0.0033 - val_loss: 0.0029 - lr: 0.0010 Epoch 17/2000 900/900 [==============================] - 1s 1ms/step - loss: 0.0027 - val_loss: 0.0028 - lr: 0.0010 Epoch 18/2000 900/900 [==============================] - 1s 1ms/step - loss: 0.0024 - val_loss: 0.0037 - lr: 0.0010 Epoch 19/2000 900/900 [==============================] - 1s 1ms/step - loss: 0.0030 - val_loss: 0.0022 - lr: 0.0010 Epoch 20/2000 900/900 [==============================] - 1s 1ms/step - loss: 0.0035 - val_loss: 0.0016 - lr: 0.0010 Epoch 21/2000 900/900 [==============================] - 1s 1ms/step - loss: 0.0013 - val_loss: 0.0039 - lr: 0.0010 Epoch 22/2000 900/900 [==============================] - 1s 1ms/step - loss: 0.0017 - val_loss: 0.0012 - lr: 0.0010 Epoch 23/2000 900/900 [==============================] - 1s 1ms/step - loss: 0.0017 - val_loss: 0.0011 - lr: 0.0010 Epoch 24/2000 900/900 [==============================] - 1s 1ms/step - loss: 0.0017 - val_loss: 0.0026 - lr: 0.0010 Epoch 25/2000 900/900 [==============================] - 1s 1ms/step - loss: 0.0010 - val_loss: 0.0046 - lr: 0.0010 Epoch 26/2000 900/900 [==============================] - 2s 2ms/step - loss: 0.0016 - val_loss: 8.2744e-04 - lr: 0.0010 Epoch 27/2000 900/900 [==============================] - 2s 2ms/step - loss: 0.0014 - val_loss: 0.0011 - lr: 0.0010 Epoch 28/2000 900/900 [==============================] - 1s 1ms/step - loss: 0.0018 - val_loss: 7.3059e-04 - lr: 0.0010 Epoch 29/2000 900/900 [==============================] - 1s 1ms/step - loss: 0.0013 - val_loss: 0.0021 - lr: 0.0010 Epoch 30/2000 900/900 [==============================] - 1s 1ms/step - loss: 0.0013 - val_loss: 5.2872e-04 - lr: 0.0010 Epoch 31/2000 900/900 [==============================] - 1s 1ms/step - loss: 9.0995e-04 - val_loss: 7.3022e-04 - lr: 0.0010 Epoch 32/2000 900/900 [==============================] - 1s 1ms/step - loss: 0.0019 - val_loss: 0.0014 - lr: 0.0010 Epoch 33/2000 900/900 [==============================] - 1s 1ms/step - loss: 6.4458e-04 - val_loss: 9.2240e-04 - lr: 0.0010 Epoch 34/2000 900/900 [==============================] - 1s 1ms/step - loss: 0.0019 - val_loss: 7.6538e-04 - lr: 0.0010 Epoch 35/2000 900/900 [==============================] - 1s 1ms/step - loss: 9.7375e-04 - val_loss: 5.7007e-04 - lr: 0.0010 Epoch 36/2000 900/900 [==============================] - 1s 1ms/step - loss: 1.5156e-04 - val_loss: 3.2450e-04 - lr: 1.0000e-04 Epoch 37/2000 900/900 [==============================] - 1s 1ms/step - loss: 1.4211e-04 - val_loss: 3.0920e-04 - lr: 1.0000e-04 Epoch 38/2000 900/900 [==============================] - 1s 1ms/step - loss: 1.3987e-04 - val_loss: 3.2047e-04 - lr: 1.0000e-04 Epoch 39/2000 900/900 [==============================] - 1s 1ms/step - loss: 1.4100e-04 - val_loss: 2.8895e-04 - lr: 1.0000e-04 Epoch 40/2000 900/900 [==============================] - 1s 1ms/step - loss: 1.3653e-04 - val_loss: 3.0373e-04 - lr: 1.0000e-04 Epoch 41/2000 900/900 [==============================] - 1s 1ms/step - loss: 1.3535e-04 - val_loss: 2.6592e-04 - lr: 1.0000e-04 Epoch 42/2000 900/900 [==============================] - 1s 1ms/step - loss: 1.3654e-04 - val_loss: 2.6799e-04 - lr: 1.0000e-04 Epoch 43/2000 900/900 [==============================] - 1s 1ms/step - loss: 1.3816e-04 - val_loss: 2.5570e-04 - lr: 1.0000e-04 Epoch 44/2000 900/900 [==============================] - 1s 1ms/step - loss: 1.2641e-04 - val_loss: 2.9339e-04 - lr: 1.0000e-04 Epoch 45/2000 900/900 [==============================] - 1s 1ms/step - loss: 1.2121e-04 - val_loss: 3.2909e-04 - lr: 1.0000e-04 Epoch 46/2000 900/900 [==============================] - 1s 1ms/step - loss: 1.2684e-04 - val_loss: 2.3325e-04 - lr: 1.0000e-04 Epoch 47/2000 900/900 [==============================] - 1s 1ms/step - loss: 1.1775e-04 - val_loss: 2.6653e-04 - lr: 1.0000e-04 Epoch 48/2000 900/900 [==============================] - 1s 1ms/step - loss: 1.1174e-04 - val_loss: 2.3168e-04 - lr: 1.0000e-04 Epoch 49/2000 900/900 [==============================] - 1s 1ms/step - loss: 1.1471e-04 - val_loss: 2.5295e-04 - lr: 1.0000e-04 Epoch 50/2000 900/900 [==============================] - 1s 1ms/step - loss: 1.1344e-04 - val_loss: 2.6298e-04 - lr: 1.0000e-04 Epoch 51/2000 900/900 [==============================] - 1s 1ms/step - loss: 1.0876e-04 - val_loss: 2.0776e-04 - lr: 1.0000e-04 Epoch 52/2000 900/900 [==============================] - 1s 1ms/step - loss: 1.0718e-04 - val_loss: 2.1121e-04 - lr: 1.0000e-04 Epoch 53/2000 900/900 [==============================] - 1s 1ms/step - loss: 1.0307e-04 - val_loss: 2.0588e-04 - lr: 1.0000e-04 Epoch 54/2000 900/900 [==============================] - 1s 1ms/step - loss: 1.0388e-04 - val_loss: 1.9177e-04 - lr: 1.0000e-04 Epoch 55/2000 900/900 [==============================] - 1s 1ms/step - loss: 1.0178e-04 - val_loss: 2.0274e-04 - lr: 1.0000e-04 Epoch 56/2000 900/900 [==============================] - 1s 1ms/step - loss: 1.0178e-04 - val_loss: 2.2856e-04 - lr: 1.0000e-04 Epoch 57/2000 900/900 [==============================] - 1s 1ms/step - loss: 1.0126e-04 - val_loss: 2.3026e-04 - lr: 1.0000e-04 Epoch 58/2000 900/900 [==============================] - 1s 1ms/step - loss: 9.3505e-05 - val_loss: 2.0378e-04 - lr: 1.0000e-04 Epoch 59/2000 900/900 [==============================] - 1s 1ms/step - loss: 9.6313e-05 - val_loss: 1.8932e-04 - lr: 1.0000e-04 Epoch 60/2000 900/900 [==============================] - 1s 1ms/step - loss: 9.1047e-05 - val_loss: 1.8083e-04 - lr: 1.0000e-04 Epoch 61/2000 900/900 [==============================] - 1s 1ms/step - loss: 9.8763e-05 - val_loss: 1.8094e-04 - lr: 1.0000e-04 Epoch 62/2000 900/900 [==============================] - 1s 1ms/step - loss: 9.1101e-05 - val_loss: 1.8211e-04 - lr: 1.0000e-04 Epoch 63/2000 900/900 [==============================] - 1s 1ms/step - loss: 9.0412e-05 - val_loss: 1.9709e-04 - lr: 1.0000e-04 Epoch 64/2000 900/900 [==============================] - 1s 1ms/step - loss: 9.3721e-05 - val_loss: 2.6898e-04 - lr: 1.0000e-04 Epoch 65/2000 900/900 [==============================] - 1s 1ms/step - loss: 9.3494e-05 - val_loss: 1.8946e-04 - lr: 1.0000e-04 Epoch 66/2000 900/900 [==============================] - 1s 1ms/step - loss: 9.1262e-05 - val_loss: 1.7357e-04 - lr: 1.0000e-04 Epoch 67/2000 900/900 [==============================] - 1s 1ms/step - loss: 8.6589e-05 - val_loss: 1.6051e-04 - lr: 1.0000e-04 Epoch 68/2000 900/900 [==============================] - 1s 1ms/step - loss: 8.8745e-05 - val_loss: 1.6204e-04 - lr: 1.0000e-04 Epoch 69/2000 900/900 [==============================] - 1s 1ms/step - loss: 8.2716e-05 - val_loss: 1.5193e-04 - lr: 1.0000e-04 Epoch 70/2000 900/900 [==============================] - 1s 1ms/step - loss: 8.5516e-05 - val_loss: 1.4889e-04 - lr: 1.0000e-04 Epoch 71/2000 900/900 [==============================] - 1s 1ms/step - loss: 8.7297e-05 - val_loss: 1.4578e-04 - lr: 1.0000e-04 Epoch 72/2000 900/900 [==============================] - 1s 1ms/step - loss: 8.0910e-05 - val_loss: 1.4944e-04 - lr: 1.0000e-04 Epoch 73/2000 900/900 [==============================] - 1s 1ms/step - loss: 8.8776e-05 - val_loss: 1.7773e-04 - lr: 1.0000e-04 Epoch 74/2000 900/900 [==============================] - 1s 1ms/step - loss: 8.2522e-05 - val_loss: 1.4596e-04 - lr: 1.0000e-04 Epoch 75/2000 900/900 [==============================] - 1s 1ms/step - loss: 8.4255e-05 - val_loss: 1.3306e-04 - lr: 1.0000e-04 Epoch 76/2000 900/900 [==============================] - 1s 1ms/step - loss: 7.8852e-05 - val_loss: 1.6480e-04 - lr: 1.0000e-04 Epoch 77/2000 900/900 [==============================] - 1s 1ms/step - loss: 9.2539e-05 - val_loss: 1.3074e-04 - lr: 1.0000e-04 Epoch 78/2000 900/900 [==============================] - 1s 1ms/step - loss: 7.7328e-05 - val_loss: 1.2497e-04 - lr: 1.0000e-04 Epoch 79/2000 900/900 [==============================] - 1s 1ms/step - loss: 9.1113e-05 - val_loss: 1.4063e-04 - lr: 1.0000e-04 Epoch 80/2000 900/900 [==============================] - 1s 1ms/step - loss: 7.1518e-05 - val_loss: 1.1932e-04 - lr: 1.0000e-04 Epoch 81/2000 900/900 [==============================] - 1s 1ms/step - loss: 6.8567e-05 - val_loss: 1.3577e-04 - lr: 1.0000e-04 Epoch 82/2000 900/900 [==============================] - 1s 1ms/step - loss: 7.6544e-05 - val_loss: 1.2487e-04 - lr: 1.0000e-04 Epoch 83/2000 900/900 [==============================] - 1s 1ms/step - loss: 7.4925e-05 - val_loss: 1.4861e-04 - lr: 1.0000e-04 Epoch 84/2000 900/900 [==============================] - 1s 1ms/step - loss: 7.4571e-05 - val_loss: 1.4264e-04 - lr: 1.0000e-04 Epoch 85/2000 900/900 [==============================] - 1s 1ms/step - loss: 6.9797e-05 - val_loss: 1.1468e-04 - lr: 1.0000e-04 Epoch 86/2000 900/900 [==============================] - 1s 1ms/step - loss: 7.4416e-05 - val_loss: 1.3066e-04 - lr: 1.0000e-04 Epoch 87/2000 900/900 [==============================] - 1s 1ms/step - loss: 7.3910e-05 - val_loss: 1.5394e-04 - lr: 1.0000e-04 Epoch 88/2000 900/900 [==============================] - 1s 1ms/step - loss: 6.9360e-05 - val_loss: 1.1626e-04 - lr: 1.0000e-04 Epoch 89/2000 900/900 [==============================] - 1s 1ms/step - loss: 6.7308e-05 - val_loss: 1.1188e-04 - lr: 1.0000e-04 Epoch 90/2000 900/900 [==============================] - 1s 1ms/step - loss: 7.2721e-05 - val_loss: 1.1300e-04 - lr: 1.0000e-04 Epoch 91/2000 900/900 [==============================] - 1s 1ms/step - loss: 6.7185e-05 - val_loss: 1.2724e-04 - lr: 1.0000e-04 Epoch 92/2000 900/900 [==============================] - 1s 1ms/step - loss: 7.7134e-05 - val_loss: 1.1569e-04 - lr: 1.0000e-04 Epoch 93/2000 900/900 [==============================] - 1s 1ms/step - loss: 6.8117e-05 - val_loss: 1.1088e-04 - lr: 1.0000e-04 Epoch 94/2000 900/900 [==============================] - 1s 1ms/step - loss: 6.8846e-05 - val_loss: 1.3633e-04 - lr: 1.0000e-04 Epoch 95/2000 900/900 [==============================] - 1s 1ms/step - loss: 6.7340e-05 - val_loss: 1.2110e-04 - lr: 1.0000e-04 Epoch 96/2000 900/900 [==============================] - 1s 1ms/step - loss: 6.7974e-05 - val_loss: 1.7295e-04 - lr: 1.0000e-04 Epoch 97/2000 900/900 [==============================] - 1s 1ms/step - loss: 6.7266e-05 - val_loss: 1.1850e-04 - lr: 1.0000e-04 Epoch 98/2000 900/900 [==============================] - 1s 1ms/step - loss: 6.2741e-05 - val_loss: 1.2930e-04 - lr: 1.0000e-04 Epoch 99/2000 900/900 [==============================] - 1s 1ms/step - loss: 7.4239e-05 - val_loss: 1.0425e-04 - lr: 1.0000e-04 Epoch 100/2000 900/900 [==============================] - 1s 1ms/step - loss: 6.8227e-05 - val_loss: 1.0128e-04 - lr: 1.0000e-04 Epoch 101/2000 900/900 [==============================] - 1s 1ms/step - loss: 6.1065e-05 - val_loss: 1.0580e-04 - lr: 1.0000e-04 Epoch 102/2000 900/900 [==============================] - 1s 1ms/step - loss: 7.2595e-05 - val_loss: 1.0268e-04 - lr: 1.0000e-04 Epoch 103/2000 900/900 [==============================] - 1s 1ms/step - loss: 6.0114e-05 - val_loss: 1.3312e-04 - lr: 1.0000e-04 Epoch 104/2000 900/900 [==============================] - 1s 2ms/step - loss: 6.9939e-05 - val_loss: 1.1080e-04 - lr: 1.0000e-04 Epoch 105/2000 900/900 [==============================] - 2s 2ms/step - loss: 6.1671e-05 - val_loss: 9.3537e-05 - lr: 1.0000e-04 Epoch 106/2000 900/900 [==============================] - 2s 2ms/step - loss: 6.9090e-05 - val_loss: 1.0073e-04 - lr: 1.0000e-04 Epoch 107/2000 900/900 [==============================] - 1s 1ms/step - loss: 6.0073e-05 - val_loss: 1.4321e-04 - lr: 1.0000e-04 Epoch 108/2000 900/900 [==============================] - 1s 1ms/step - loss: 5.8665e-05 - val_loss: 1.0323e-04 - lr: 1.0000e-04 Epoch 109/2000 900/900 [==============================] - 1s 1ms/step - loss: 6.1377e-05 - val_loss: 2.0407e-04 - lr: 1.0000e-04 Epoch 110/2000 900/900 [==============================] - 1s 1ms/step - loss: 6.5973e-05 - val_loss: 9.2555e-05 - lr: 1.0000e-04 Epoch 111/2000 900/900 [==============================] - 1s 1ms/step - loss: 6.1832e-05 - val_loss: 1.1252e-04 - lr: 1.0000e-04 Epoch 112/2000 900/900 [==============================] - 1s 1ms/step - loss: 6.1293e-05 - val_loss: 8.5153e-05 - lr: 1.0000e-04 Epoch 113/2000 900/900 [==============================] - 1s 1ms/step - loss: 5.6221e-05 - val_loss: 9.7618e-05 - lr: 1.0000e-04 Epoch 114/2000 900/900 [==============================] - 1s 1ms/step - loss: 6.3836e-05 - val_loss: 9.4311e-05 - lr: 1.0000e-04 Epoch 115/2000 900/900 [==============================] - 1s 1ms/step - loss: 5.8413e-05 - val_loss: 1.3108e-04 - lr: 1.0000e-04 Epoch 116/2000 900/900 [==============================] - 1s 1ms/step - loss: 6.1561e-05 - val_loss: 9.7013e-05 - lr: 1.0000e-04 Epoch 117/2000 900/900 [==============================] - 1s 1ms/step - loss: 6.2772e-05 - val_loss: 1.0615e-04 - lr: 1.0000e-04 Epoch 118/2000 900/900 [==============================] - 1s 1ms/step - loss: 5.7437e-05 - val_loss: 1.3355e-04 - lr: 1.0000e-04 Epoch 119/2000 900/900 [==============================] - 1s 1ms/step - loss: 5.7079e-05 - val_loss: 9.6611e-05 - lr: 1.0000e-04 Epoch 120/2000 900/900 [==============================] - 1s 1ms/step - loss: 6.2548e-05 - val_loss: 1.0475e-04 - lr: 1.0000e-04 Epoch 121/2000 900/900 [==============================] - 1s 1ms/step - loss: 5.7250e-05 - val_loss: 1.1300e-04 - lr: 1.0000e-04 Epoch 122/2000 900/900 [==============================] - 1s 1ms/step - loss: 6.4314e-05 - val_loss: 9.1217e-05 - lr: 1.0000e-04
# plotting loss function
fig, ax = plt.subplots(figsize=(10, 7))
ax.plot(history.history['loss'], label='Training Loss')
ax.plot(history.history['val_loss'], label='Validation Loss')
ax.set(
title='Processo de otimização das funções de perda',
ylabel='Loss',
xlabel='Epoch'
)
plt.legend()
plt.tight_layout()
plt.show()
# true points and predicted points
fig, ax = plt.subplots(figsize=(15, 9), subplot_kw=dict(projection='3d'))
ax.plot_wireframe(x1, x2, y, linewidths=0.5, color='lightgrey')
ax.scatter(x_test[:,0], x_test[:,1], y_test, s=14, color='C0', label='Training data')
ax.scatter(x_test[:,0], x_test[:,1], mlp.predict(x_test), s=15, marker='^', color='C1', label='Test data')
ax.set(
xlabel='$x_1$',
ylabel='$x_2$',
zlabel='$f(x_1, x_2)$'
)
plt.legend()
plt.tight_layout()
plt.show()
63/63 [==============================] - 0s 1ms/step
mse = mean_squared_error(y_test, mlp.predict(x_test))
rmse = mean_squared_error(y_test, mlp.predict(x_test), squared=False)
mae = mean_absolute_error(y_test, mlp.predict(x_test))
63/63 [==============================] - 0s 903us/step 63/63 [==============================] - 0s 988us/step 63/63 [==============================] - 0s 763us/step
print(f"Mean Squared Error: {mse}")
print(f"Root Mean Squared Error: {rmse}")
print(f"Mean Absolute Error: {mae}")
Mean Squared Error: 4.9327674554969855e-05 Root Mean Squared Error: 0.007023366326411421 Mean Absolute Error: 0.0053168037015980385
Considere o problema de classificação de padrões bidimensionais constituído neste caso de 5 padrões. A distribuição dos padrões tem como base um quadrado centrado na origem interceptando os eixos nos pontos +1 e -1 de cada eixo. Os pontos +1 e -1 de cada eixo são centros de quatro semicírculos que se interceptam no interior do quadrado originando as classes 1,2,3,4 e a outra classe corresponde as regiões de não interseção. Após gerar aleato- riamente os dados que venham formar estas distribuições de dados, selecione um conjunto de treinamento e um conjunto de validação com o rótulo de cada classe. Solucione este problema considerando uma rede perceptron de múltiplas camada. Apresente na solução a curva do erro médio de treinamento e a curva do erro médio de teste. Apresente também a matriz de confusão.
# Semicírculo azul
def c1(x, y):
if (x + 1)**2 + y**2 <= 1:
return 1
else:
return 0
#Sermicírculo verde
def c2(x, y):
if (x - 1)**2 + y**2 <= 1:
return 1
else:
return 0
#Semicírculo amarelo
def c3(x, y):
if x**2 + (y + 1)**2 <= 1:
return 1
else:
return 0
#Semicírculo vermelha=o
def c4(x, y):
if x**2 + (y - 1)**2 <= 1:
return 1
else:
return 0
x, y = np.meshgrid(np.linspace(-1, 1, 100), np.linspace(-1, 1, 100))
points = np.vstack(list(zip(x.ravel(), y.ravel())))
lista = []
for x_i, y_i in points:
if c1(x_i, y_i) + c4(x_i, y_i) == 2:
lista.append(1)
elif c2(x_i, y_i) + c4(x_i, y_i) == 2:
lista.append(2)
elif c2(x_i, y_i) + c3(x_i, y_i) == 2:
lista.append(3)
elif c1(x_i, y_i) + c3(x_i, y_i) == 2:
lista.append(4)
else:
lista.append(0)
labels = np.array(lista)
x_train, x_test, y_train, y_test = train_test_split(points, labels, test_size=0.2, stratify=labels)
print(f"shape x train: {x_train.shape}")
print(f"shape y train: {y_train.shape}")
print(f"shape x test: {x_test.shape}")
print(f"shape y test: {y_test.shape}")
shape x train: (8000, 2) shape y train: (8000,) shape x test: (2000, 2) shape y test: (2000,)
fig, ax = plt.subplots(ncols=3, figsize=(40, 12))
color1 = (181/255, 181/255, 181/255, 1.0)
color2 = (38/255, 118/255, 222/255, 1.0)
color3 = (38/255, 222/255, 118/255, 1.0)
color4 = (235/255, 91/255, 156/255, 1.0)
color5 = (240/255, 130/255, 44/255, 1.0)
colormap = np.array([color1, color2, color3, color4, color5])
dataset_scatter = ax[0].scatter(points[:,0], points[:,1], c=colormap[labels], marker='d')
ax[0].set(
title='Dataset',
xlabel='$x$',
ylabel='$y$'
)
ax[1].scatter(x_train[:,0], x_train[:,1], c=colormap[y_train], marker='d')
ax[1].set(
title='Training set',
xlabel='$x$',
ylabel='$y$'
)
ax[2].scatter(x_test[:,0], x_test[:,1], c=colormap[y_test], marker='d')
ax[2].set(
title='Test set',
xlabel='$x$',
ylabel='$y$'
)
from matplotlib.patches import Patch
from matplotlib.lines import Line2D
legend_elements = [
Line2D([0], [0], marker='o', color='w', label='Scatter', markerfacecolor=(181/255, 181/255, 181/255), markersize=15),
Line2D([0], [0], marker='o', color='w', label='Scatter', markerfacecolor=(38/255, 118/255, 222/255), markersize=15),
Line2D([0], [0], marker='o', color='w', label='Scatter', markerfacecolor=(38/255, 222/255, 118/255), markersize=15),
Line2D([0], [0], marker='o', color='w', label='Scatter', markerfacecolor=(235/255, 91/255, 156/255), markersize=15),
Line2D([0], [0], marker='o', color='w', label='Scatter', markerfacecolor=(240/255, 130/255, 44/255), markersize=15)
]
fig.legend(
legend_elements,
['0', '1', '2', '3', '4'],
loc='lower center',
title='Classes'
)
plt.show()
mlp = Sequential([
Dense(64, activation='relu', input_shape=(2,)),
Dense(32, activation='relu'),
Dense(16, activation='relu'),
Dense(8, activation='relu'),
Dense(5, activation='softmax')
])
mlp.compile(
loss='sparse_categorical_crossentropy',
optimizer='adam',
metrics=['acc']
)
mlp.summary()
Model: "sequential_1"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
dense_5 (Dense) (None, 64) 192
dense_6 (Dense) (None, 32) 2080
dense_7 (Dense) (None, 16) 528
dense_8 (Dense) (None, 8) 136
dense_9 (Dense) (None, 5) 45
=================================================================
Total params: 2,981
Trainable params: 2,981
Non-trainable params: 0
_________________________________________________________________
history = mlp.fit(
x_train, y_train.reshape((-1,1)),
validation_split=0.1,
batch_size=10,
epochs=2000,
callbacks=[
tf.keras.callbacks.EarlyStopping(monitor='loss', patience=10),
tf.keras.callbacks.ReduceLROnPlateau(monitor='loss', factor=0.1, patience=5, min_lr=0.0001)
]
)
Epoch 1/2000 720/720 [==============================] - 2s 3ms/step - loss: 0.7271 - acc: 0.6861 - val_loss: 0.4313 - val_acc: 0.8350 - lr: 0.0010 Epoch 2/2000 720/720 [==============================] - 2s 3ms/step - loss: 0.3175 - acc: 0.8826 - val_loss: 0.2762 - val_acc: 0.8975 - lr: 0.0010 Epoch 3/2000 720/720 [==============================] - 2s 3ms/step - loss: 0.2327 - acc: 0.9156 - val_loss: 0.2305 - val_acc: 0.9125 - lr: 0.0010 Epoch 4/2000 720/720 [==============================] - 2s 3ms/step - loss: 0.1948 - acc: 0.9292 - val_loss: 0.1951 - val_acc: 0.9162 - lr: 0.0010 Epoch 5/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.1680 - acc: 0.9410 - val_loss: 0.1584 - val_acc: 0.9438 - lr: 0.0010 Epoch 6/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.1588 - acc: 0.9375 - val_loss: 0.1548 - val_acc: 0.9375 - lr: 0.0010 Epoch 7/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.1468 - acc: 0.9440 - val_loss: 0.1422 - val_acc: 0.9438 - lr: 0.0010 Epoch 8/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.1337 - acc: 0.9468 - val_loss: 0.1154 - val_acc: 0.9550 - lr: 0.0010 Epoch 9/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.1251 - acc: 0.9507 - val_loss: 0.1369 - val_acc: 0.9350 - lr: 0.0010 Epoch 10/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.1272 - acc: 0.9499 - val_loss: 0.1201 - val_acc: 0.9575 - lr: 0.0010 Epoch 11/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.1169 - acc: 0.9533 - val_loss: 0.1246 - val_acc: 0.9438 - lr: 0.0010 Epoch 12/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.1170 - acc: 0.9517 - val_loss: 0.1711 - val_acc: 0.9212 - lr: 0.0010 Epoch 13/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.1165 - acc: 0.9532 - val_loss: 0.1301 - val_acc: 0.9438 - lr: 0.0010 Epoch 14/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.1164 - acc: 0.9524 - val_loss: 0.0993 - val_acc: 0.9588 - lr: 0.0010 Epoch 15/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.1120 - acc: 0.9519 - val_loss: 0.1097 - val_acc: 0.9488 - lr: 0.0010 Epoch 16/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.1062 - acc: 0.9569 - val_loss: 0.1492 - val_acc: 0.9337 - lr: 0.0010 Epoch 17/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.1104 - acc: 0.9533 - val_loss: 0.0911 - val_acc: 0.9588 - lr: 0.0010 Epoch 18/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.1109 - acc: 0.9550 - val_loss: 0.1208 - val_acc: 0.9525 - lr: 0.0010 Epoch 19/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.1010 - acc: 0.9592 - val_loss: 0.1122 - val_acc: 0.9513 - lr: 0.0010 Epoch 20/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0980 - acc: 0.9599 - val_loss: 0.0863 - val_acc: 0.9638 - lr: 0.0010 Epoch 21/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0965 - acc: 0.9596 - val_loss: 0.0889 - val_acc: 0.9613 - lr: 0.0010 Epoch 22/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0994 - acc: 0.9597 - val_loss: 0.1672 - val_acc: 0.9262 - lr: 0.0010 Epoch 23/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0979 - acc: 0.9589 - val_loss: 0.1394 - val_acc: 0.9463 - lr: 0.0010 Epoch 24/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.1035 - acc: 0.9558 - val_loss: 0.0798 - val_acc: 0.9638 - lr: 0.0010 Epoch 25/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0924 - acc: 0.9633 - val_loss: 0.0743 - val_acc: 0.9688 - lr: 0.0010 Epoch 26/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0937 - acc: 0.9601 - val_loss: 0.1888 - val_acc: 0.9425 - lr: 0.0010 Epoch 27/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.1012 - acc: 0.9583 - val_loss: 0.0916 - val_acc: 0.9600 - lr: 0.0010 Epoch 28/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0925 - acc: 0.9617 - val_loss: 0.0768 - val_acc: 0.9613 - lr: 0.0010 Epoch 29/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0933 - acc: 0.9632 - val_loss: 0.1094 - val_acc: 0.9513 - lr: 0.0010 Epoch 30/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0930 - acc: 0.9617 - val_loss: 0.0828 - val_acc: 0.9613 - lr: 0.0010 Epoch 31/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0574 - acc: 0.9829 - val_loss: 0.0603 - val_acc: 0.9737 - lr: 1.0000e-04 Epoch 32/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0554 - acc: 0.9847 - val_loss: 0.0665 - val_acc: 0.9712 - lr: 1.0000e-04 Epoch 33/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0549 - acc: 0.9856 - val_loss: 0.0587 - val_acc: 0.9762 - lr: 1.0000e-04 Epoch 34/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0556 - acc: 0.9844 - val_loss: 0.0582 - val_acc: 0.9800 - lr: 1.0000e-04 Epoch 35/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0544 - acc: 0.9847 - val_loss: 0.0599 - val_acc: 0.9787 - lr: 1.0000e-04 Epoch 36/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0542 - acc: 0.9847 - val_loss: 0.0614 - val_acc: 0.9762 - lr: 1.0000e-04 Epoch 37/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0544 - acc: 0.9851 - val_loss: 0.0634 - val_acc: 0.9737 - lr: 1.0000e-04 Epoch 38/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0536 - acc: 0.9846 - val_loss: 0.0570 - val_acc: 0.9787 - lr: 1.0000e-04 Epoch 39/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0530 - acc: 0.9854 - val_loss: 0.0511 - val_acc: 0.9837 - lr: 1.0000e-04 Epoch 40/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0536 - acc: 0.9858 - val_loss: 0.0554 - val_acc: 0.9800 - lr: 1.0000e-04 Epoch 41/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0527 - acc: 0.9862 - val_loss: 0.0612 - val_acc: 0.9737 - lr: 1.0000e-04 Epoch 42/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0527 - acc: 0.9856 - val_loss: 0.0540 - val_acc: 0.9837 - lr: 1.0000e-04 Epoch 43/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0519 - acc: 0.9861 - val_loss: 0.0597 - val_acc: 0.9750 - lr: 1.0000e-04 Epoch 44/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0513 - acc: 0.9869 - val_loss: 0.0547 - val_acc: 0.9762 - lr: 1.0000e-04 Epoch 45/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0514 - acc: 0.9861 - val_loss: 0.0577 - val_acc: 0.9800 - lr: 1.0000e-04 Epoch 46/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0511 - acc: 0.9850 - val_loss: 0.0548 - val_acc: 0.9775 - lr: 1.0000e-04 Epoch 47/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0518 - acc: 0.9847 - val_loss: 0.0580 - val_acc: 0.9775 - lr: 1.0000e-04 Epoch 48/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0509 - acc: 0.9861 - val_loss: 0.0565 - val_acc: 0.9787 - lr: 1.0000e-04 Epoch 49/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0502 - acc: 0.9868 - val_loss: 0.0508 - val_acc: 0.9850 - lr: 1.0000e-04 Epoch 50/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0505 - acc: 0.9854 - val_loss: 0.0611 - val_acc: 0.9750 - lr: 1.0000e-04 Epoch 51/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0492 - acc: 0.9862 - val_loss: 0.0535 - val_acc: 0.9800 - lr: 1.0000e-04 Epoch 52/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0493 - acc: 0.9858 - val_loss: 0.0552 - val_acc: 0.9787 - lr: 1.0000e-04 Epoch 53/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0495 - acc: 0.9868 - val_loss: 0.0559 - val_acc: 0.9750 - lr: 1.0000e-04 Epoch 54/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0487 - acc: 0.9856 - val_loss: 0.0526 - val_acc: 0.9800 - lr: 1.0000e-04 Epoch 55/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0490 - acc: 0.9867 - val_loss: 0.0565 - val_acc: 0.9775 - lr: 1.0000e-04 Epoch 56/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0484 - acc: 0.9864 - val_loss: 0.0520 - val_acc: 0.9825 - lr: 1.0000e-04 Epoch 57/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0481 - acc: 0.9882 - val_loss: 0.0494 - val_acc: 0.9825 - lr: 1.0000e-04 Epoch 58/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0476 - acc: 0.9882 - val_loss: 0.0550 - val_acc: 0.9800 - lr: 1.0000e-04 Epoch 59/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0478 - acc: 0.9872 - val_loss: 0.0577 - val_acc: 0.9762 - lr: 1.0000e-04 Epoch 60/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0480 - acc: 0.9864 - val_loss: 0.0521 - val_acc: 0.9787 - lr: 1.0000e-04 Epoch 61/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0476 - acc: 0.9872 - val_loss: 0.0478 - val_acc: 0.9825 - lr: 1.0000e-04 Epoch 62/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0477 - acc: 0.9865 - val_loss: 0.0485 - val_acc: 0.9812 - lr: 1.0000e-04 Epoch 63/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0471 - acc: 0.9857 - val_loss: 0.0510 - val_acc: 0.9800 - lr: 1.0000e-04 Epoch 64/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0464 - acc: 0.9890 - val_loss: 0.0604 - val_acc: 0.9737 - lr: 1.0000e-04 Epoch 65/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0470 - acc: 0.9878 - val_loss: 0.0538 - val_acc: 0.9750 - lr: 1.0000e-04 Epoch 66/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0463 - acc: 0.9878 - val_loss: 0.0523 - val_acc: 0.9787 - lr: 1.0000e-04 Epoch 67/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0462 - acc: 0.9868 - val_loss: 0.0496 - val_acc: 0.9850 - lr: 1.0000e-04 Epoch 68/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0458 - acc: 0.9871 - val_loss: 0.0516 - val_acc: 0.9800 - lr: 1.0000e-04 Epoch 69/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0458 - acc: 0.9875 - val_loss: 0.0488 - val_acc: 0.9825 - lr: 1.0000e-04 Epoch 70/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0460 - acc: 0.9871 - val_loss: 0.0504 - val_acc: 0.9775 - lr: 1.0000e-04 Epoch 71/2000 720/720 [==============================] - 2s 3ms/step - loss: 0.0462 - acc: 0.9868 - val_loss: 0.0586 - val_acc: 0.9725 - lr: 1.0000e-04 Epoch 72/2000 720/720 [==============================] - 2s 3ms/step - loss: 0.0450 - acc: 0.9879 - val_loss: 0.0515 - val_acc: 0.9825 - lr: 1.0000e-04 Epoch 73/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0450 - acc: 0.9882 - val_loss: 0.0560 - val_acc: 0.9725 - lr: 1.0000e-04 Epoch 74/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0450 - acc: 0.9890 - val_loss: 0.0512 - val_acc: 0.9787 - lr: 1.0000e-04 Epoch 75/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0452 - acc: 0.9881 - val_loss: 0.0480 - val_acc: 0.9812 - lr: 1.0000e-04 Epoch 76/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0441 - acc: 0.9889 - val_loss: 0.0443 - val_acc: 0.9862 - lr: 1.0000e-04 Epoch 77/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0446 - acc: 0.9868 - val_loss: 0.0481 - val_acc: 0.9825 - lr: 1.0000e-04 Epoch 78/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0442 - acc: 0.9876 - val_loss: 0.0562 - val_acc: 0.9737 - lr: 1.0000e-04 Epoch 79/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0439 - acc: 0.9885 - val_loss: 0.0434 - val_acc: 0.9862 - lr: 1.0000e-04 Epoch 80/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0443 - acc: 0.9867 - val_loss: 0.0488 - val_acc: 0.9812 - lr: 1.0000e-04 Epoch 81/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0437 - acc: 0.9893 - val_loss: 0.0470 - val_acc: 0.9812 - lr: 1.0000e-04 Epoch 82/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0436 - acc: 0.9885 - val_loss: 0.0456 - val_acc: 0.9875 - lr: 1.0000e-04 Epoch 83/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0432 - acc: 0.9885 - val_loss: 0.0515 - val_acc: 0.9775 - lr: 1.0000e-04 Epoch 84/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0443 - acc: 0.9871 - val_loss: 0.0486 - val_acc: 0.9800 - lr: 1.0000e-04 Epoch 85/2000 720/720 [==============================] - 2s 3ms/step - loss: 0.0431 - acc: 0.9886 - val_loss: 0.0446 - val_acc: 0.9875 - lr: 1.0000e-04 Epoch 86/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0430 - acc: 0.9886 - val_loss: 0.0506 - val_acc: 0.9762 - lr: 1.0000e-04 Epoch 87/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0430 - acc: 0.9886 - val_loss: 0.0503 - val_acc: 0.9800 - lr: 1.0000e-04 Epoch 88/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0430 - acc: 0.9890 - val_loss: 0.0529 - val_acc: 0.9762 - lr: 1.0000e-04 Epoch 89/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0423 - acc: 0.9887 - val_loss: 0.0550 - val_acc: 0.9712 - lr: 1.0000e-04 Epoch 90/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0426 - acc: 0.9889 - val_loss: 0.0499 - val_acc: 0.9775 - lr: 1.0000e-04 Epoch 91/2000 720/720 [==============================] - 3s 4ms/step - loss: 0.0415 - acc: 0.9906 - val_loss: 0.0428 - val_acc: 0.9862 - lr: 1.0000e-04 Epoch 92/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0422 - acc: 0.9882 - val_loss: 0.0472 - val_acc: 0.9837 - lr: 1.0000e-04 Epoch 93/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0422 - acc: 0.9883 - val_loss: 0.0434 - val_acc: 0.9825 - lr: 1.0000e-04 Epoch 94/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0425 - acc: 0.9883 - val_loss: 0.0464 - val_acc: 0.9837 - lr: 1.0000e-04 Epoch 95/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0415 - acc: 0.9903 - val_loss: 0.0567 - val_acc: 0.9762 - lr: 1.0000e-04 Epoch 96/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0416 - acc: 0.9903 - val_loss: 0.0423 - val_acc: 0.9850 - lr: 1.0000e-04 Epoch 97/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0418 - acc: 0.9889 - val_loss: 0.0464 - val_acc: 0.9750 - lr: 1.0000e-04 Epoch 98/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0410 - acc: 0.9906 - val_loss: 0.0502 - val_acc: 0.9750 - lr: 1.0000e-04 Epoch 99/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0414 - acc: 0.9896 - val_loss: 0.0430 - val_acc: 0.9887 - lr: 1.0000e-04 Epoch 100/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0408 - acc: 0.9894 - val_loss: 0.0437 - val_acc: 0.9875 - lr: 1.0000e-04 Epoch 101/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0413 - acc: 0.9885 - val_loss: 0.0478 - val_acc: 0.9800 - lr: 1.0000e-04 Epoch 102/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0408 - acc: 0.9893 - val_loss: 0.0418 - val_acc: 0.9862 - lr: 1.0000e-04 Epoch 103/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0403 - acc: 0.9896 - val_loss: 0.0454 - val_acc: 0.9787 - lr: 1.0000e-04 Epoch 104/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0406 - acc: 0.9881 - val_loss: 0.0420 - val_acc: 0.9887 - lr: 1.0000e-04 Epoch 105/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0410 - acc: 0.9883 - val_loss: 0.0502 - val_acc: 0.9762 - lr: 1.0000e-04 Epoch 106/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0410 - acc: 0.9889 - val_loss: 0.0394 - val_acc: 0.9900 - lr: 1.0000e-04 Epoch 107/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0401 - acc: 0.9887 - val_loss: 0.0515 - val_acc: 0.9762 - lr: 1.0000e-04 Epoch 108/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0400 - acc: 0.9900 - val_loss: 0.0409 - val_acc: 0.9850 - lr: 1.0000e-04 Epoch 109/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0405 - acc: 0.9892 - val_loss: 0.0411 - val_acc: 0.9862 - lr: 1.0000e-04 Epoch 110/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0407 - acc: 0.9879 - val_loss: 0.0589 - val_acc: 0.9688 - lr: 1.0000e-04 Epoch 111/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0392 - acc: 0.9896 - val_loss: 0.0581 - val_acc: 0.9762 - lr: 1.0000e-04 Epoch 112/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0394 - acc: 0.9896 - val_loss: 0.0492 - val_acc: 0.9800 - lr: 1.0000e-04 Epoch 113/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0395 - acc: 0.9890 - val_loss: 0.0449 - val_acc: 0.9837 - lr: 1.0000e-04 Epoch 114/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0396 - acc: 0.9878 - val_loss: 0.0389 - val_acc: 0.9875 - lr: 1.0000e-04 Epoch 115/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0393 - acc: 0.9901 - val_loss: 0.0464 - val_acc: 0.9750 - lr: 1.0000e-04 Epoch 116/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0395 - acc: 0.9893 - val_loss: 0.0502 - val_acc: 0.9725 - lr: 1.0000e-04 Epoch 117/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0390 - acc: 0.9885 - val_loss: 0.0448 - val_acc: 0.9800 - lr: 1.0000e-04 Epoch 118/2000 720/720 [==============================] - 2s 3ms/step - loss: 0.0385 - acc: 0.9899 - val_loss: 0.0396 - val_acc: 0.9862 - lr: 1.0000e-04 Epoch 119/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0391 - acc: 0.9900 - val_loss: 0.0409 - val_acc: 0.9837 - lr: 1.0000e-04 Epoch 120/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0386 - acc: 0.9914 - val_loss: 0.0514 - val_acc: 0.9775 - lr: 1.0000e-04 Epoch 121/2000 720/720 [==============================] - 2s 3ms/step - loss: 0.0392 - acc: 0.9889 - val_loss: 0.0431 - val_acc: 0.9812 - lr: 1.0000e-04 Epoch 122/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0389 - acc: 0.9894 - val_loss: 0.0447 - val_acc: 0.9775 - lr: 1.0000e-04 Epoch 123/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0382 - acc: 0.9894 - val_loss: 0.0507 - val_acc: 0.9750 - lr: 1.0000e-04 Epoch 124/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0384 - acc: 0.9899 - val_loss: 0.0406 - val_acc: 0.9862 - lr: 1.0000e-04 Epoch 125/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0378 - acc: 0.9901 - val_loss: 0.0470 - val_acc: 0.9825 - lr: 1.0000e-04 Epoch 126/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0384 - acc: 0.9889 - val_loss: 0.0460 - val_acc: 0.9812 - lr: 1.0000e-04 Epoch 127/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0383 - acc: 0.9889 - val_loss: 0.0390 - val_acc: 0.9887 - lr: 1.0000e-04 Epoch 128/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0381 - acc: 0.9875 - val_loss: 0.0396 - val_acc: 0.9862 - lr: 1.0000e-04 Epoch 129/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0381 - acc: 0.9894 - val_loss: 0.0554 - val_acc: 0.9750 - lr: 1.0000e-04 Epoch 130/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0378 - acc: 0.9889 - val_loss: 0.0478 - val_acc: 0.9725 - lr: 1.0000e-04 Epoch 131/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0373 - acc: 0.9903 - val_loss: 0.0416 - val_acc: 0.9850 - lr: 1.0000e-04 Epoch 132/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0383 - acc: 0.9893 - val_loss: 0.0394 - val_acc: 0.9900 - lr: 1.0000e-04 Epoch 133/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0372 - acc: 0.9899 - val_loss: 0.0383 - val_acc: 0.9900 - lr: 1.0000e-04 Epoch 134/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0373 - acc: 0.9901 - val_loss: 0.0469 - val_acc: 0.9775 - lr: 1.0000e-04 Epoch 135/2000 720/720 [==============================] - 2s 3ms/step - loss: 0.0371 - acc: 0.9897 - val_loss: 0.0401 - val_acc: 0.9850 - lr: 1.0000e-04 Epoch 136/2000 720/720 [==============================] - 2s 3ms/step - loss: 0.0377 - acc: 0.9890 - val_loss: 0.0436 - val_acc: 0.9800 - lr: 1.0000e-04 Epoch 137/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0368 - acc: 0.9894 - val_loss: 0.0405 - val_acc: 0.9825 - lr: 1.0000e-04 Epoch 138/2000 720/720 [==============================] - 2s 3ms/step - loss: 0.0363 - acc: 0.9897 - val_loss: 0.0450 - val_acc: 0.9800 - lr: 1.0000e-04 Epoch 139/2000 720/720 [==============================] - 2s 3ms/step - loss: 0.0371 - acc: 0.9900 - val_loss: 0.0392 - val_acc: 0.9862 - lr: 1.0000e-04 Epoch 140/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0370 - acc: 0.9899 - val_loss: 0.0474 - val_acc: 0.9750 - lr: 1.0000e-04 Epoch 141/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0363 - acc: 0.9892 - val_loss: 0.0386 - val_acc: 0.9812 - lr: 1.0000e-04 Epoch 142/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0371 - acc: 0.9889 - val_loss: 0.0436 - val_acc: 0.9850 - lr: 1.0000e-04 Epoch 143/2000 720/720 [==============================] - 2s 3ms/step - loss: 0.0365 - acc: 0.9908 - val_loss: 0.0487 - val_acc: 0.9737 - lr: 1.0000e-04 Epoch 144/2000 720/720 [==============================] - 2s 3ms/step - loss: 0.0366 - acc: 0.9892 - val_loss: 0.0380 - val_acc: 0.9850 - lr: 1.0000e-04 Epoch 145/2000 720/720 [==============================] - 2s 3ms/step - loss: 0.0363 - acc: 0.9900 - val_loss: 0.0393 - val_acc: 0.9837 - lr: 1.0000e-04 Epoch 146/2000 720/720 [==============================] - 2s 3ms/step - loss: 0.0371 - acc: 0.9899 - val_loss: 0.0374 - val_acc: 0.9887 - lr: 1.0000e-04 Epoch 147/2000 720/720 [==============================] - 2s 3ms/step - loss: 0.0362 - acc: 0.9894 - val_loss: 0.0489 - val_acc: 0.9787 - lr: 1.0000e-04 Epoch 148/2000 720/720 [==============================] - 2s 3ms/step - loss: 0.0359 - acc: 0.9894 - val_loss: 0.0459 - val_acc: 0.9775 - lr: 1.0000e-04 Epoch 149/2000 720/720 [==============================] - 2s 3ms/step - loss: 0.0360 - acc: 0.9899 - val_loss: 0.0328 - val_acc: 0.9937 - lr: 1.0000e-04 Epoch 150/2000 720/720 [==============================] - 2s 3ms/step - loss: 0.0357 - acc: 0.9892 - val_loss: 0.0459 - val_acc: 0.9800 - lr: 1.0000e-04 Epoch 151/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0351 - acc: 0.9896 - val_loss: 0.0425 - val_acc: 0.9850 - lr: 1.0000e-04 Epoch 152/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0365 - acc: 0.9885 - val_loss: 0.0402 - val_acc: 0.9800 - lr: 1.0000e-04 Epoch 153/2000 720/720 [==============================] - 2s 3ms/step - loss: 0.0350 - acc: 0.9906 - val_loss: 0.0390 - val_acc: 0.9850 - lr: 1.0000e-04 Epoch 154/2000 720/720 [==============================] - 2s 3ms/step - loss: 0.0362 - acc: 0.9894 - val_loss: 0.0378 - val_acc: 0.9825 - lr: 1.0000e-04 Epoch 155/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0365 - acc: 0.9878 - val_loss: 0.0366 - val_acc: 0.9900 - lr: 1.0000e-04 Epoch 156/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0352 - acc: 0.9907 - val_loss: 0.0386 - val_acc: 0.9862 - lr: 1.0000e-04 Epoch 157/2000 720/720 [==============================] - 2s 3ms/step - loss: 0.0351 - acc: 0.9897 - val_loss: 0.0435 - val_acc: 0.9762 - lr: 1.0000e-04 Epoch 158/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0350 - acc: 0.9910 - val_loss: 0.0530 - val_acc: 0.9700 - lr: 1.0000e-04 Epoch 159/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0345 - acc: 0.9904 - val_loss: 0.0379 - val_acc: 0.9812 - lr: 1.0000e-04 Epoch 160/2000 720/720 [==============================] - 2s 3ms/step - loss: 0.0359 - acc: 0.9886 - val_loss: 0.0375 - val_acc: 0.9887 - lr: 1.0000e-04 Epoch 161/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0347 - acc: 0.9906 - val_loss: 0.0394 - val_acc: 0.9800 - lr: 1.0000e-04 Epoch 162/2000 720/720 [==============================] - 3s 4ms/step - loss: 0.0345 - acc: 0.9900 - val_loss: 0.0400 - val_acc: 0.9875 - lr: 1.0000e-04 Epoch 163/2000 720/720 [==============================] - 3s 4ms/step - loss: 0.0346 - acc: 0.9907 - val_loss: 0.0372 - val_acc: 0.9850 - lr: 1.0000e-04 Epoch 164/2000 720/720 [==============================] - 2s 3ms/step - loss: 0.0348 - acc: 0.9892 - val_loss: 0.0340 - val_acc: 0.9925 - lr: 1.0000e-04 Epoch 165/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0346 - acc: 0.9897 - val_loss: 0.0386 - val_acc: 0.9850 - lr: 1.0000e-04 Epoch 166/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0356 - acc: 0.9887 - val_loss: 0.0398 - val_acc: 0.9825 - lr: 1.0000e-04 Epoch 167/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0347 - acc: 0.9904 - val_loss: 0.0436 - val_acc: 0.9825 - lr: 1.0000e-04 Epoch 168/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0345 - acc: 0.9897 - val_loss: 0.0350 - val_acc: 0.9875 - lr: 1.0000e-04 Epoch 169/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0333 - acc: 0.9899 - val_loss: 0.0433 - val_acc: 0.9787 - lr: 1.0000e-04 Epoch 170/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0345 - acc: 0.9890 - val_loss: 0.0370 - val_acc: 0.9837 - lr: 1.0000e-04 Epoch 171/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0343 - acc: 0.9897 - val_loss: 0.0417 - val_acc: 0.9825 - lr: 1.0000e-04 Epoch 172/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0342 - acc: 0.9896 - val_loss: 0.0353 - val_acc: 0.9862 - lr: 1.0000e-04 Epoch 173/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0348 - acc: 0.9904 - val_loss: 0.0407 - val_acc: 0.9812 - lr: 1.0000e-04 Epoch 174/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0334 - acc: 0.9912 - val_loss: 0.0405 - val_acc: 0.9812 - lr: 1.0000e-04 Epoch 175/2000 720/720 [==============================] - 3s 4ms/step - loss: 0.0338 - acc: 0.9912 - val_loss: 0.0392 - val_acc: 0.9825 - lr: 1.0000e-04 Epoch 176/2000 720/720 [==============================] - 4s 6ms/step - loss: 0.0341 - acc: 0.9892 - val_loss: 0.0326 - val_acc: 0.9887 - lr: 1.0000e-04 Epoch 177/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0339 - acc: 0.9907 - val_loss: 0.0433 - val_acc: 0.9775 - lr: 1.0000e-04 Epoch 178/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0328 - acc: 0.9896 - val_loss: 0.0374 - val_acc: 0.9850 - lr: 1.0000e-04 Epoch 179/2000 720/720 [==============================] - 5s 8ms/step - loss: 0.0339 - acc: 0.9889 - val_loss: 0.0330 - val_acc: 0.9887 - lr: 1.0000e-04 Epoch 180/2000 720/720 [==============================] - 3s 4ms/step - loss: 0.0330 - acc: 0.9907 - val_loss: 0.0345 - val_acc: 0.9912 - lr: 1.0000e-04 Epoch 181/2000 720/720 [==============================] - 4s 6ms/step - loss: 0.0335 - acc: 0.9903 - val_loss: 0.0344 - val_acc: 0.9887 - lr: 1.0000e-04 Epoch 182/2000 720/720 [==============================] - 4s 5ms/step - loss: 0.0325 - acc: 0.9915 - val_loss: 0.0348 - val_acc: 0.9875 - lr: 1.0000e-04 Epoch 183/2000 720/720 [==============================] - 2s 3ms/step - loss: 0.0334 - acc: 0.9904 - val_loss: 0.0411 - val_acc: 0.9787 - lr: 1.0000e-04 Epoch 184/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0336 - acc: 0.9906 - val_loss: 0.0375 - val_acc: 0.9837 - lr: 1.0000e-04 Epoch 185/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0334 - acc: 0.9903 - val_loss: 0.0397 - val_acc: 0.9825 - lr: 1.0000e-04 Epoch 186/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0329 - acc: 0.9907 - val_loss: 0.0452 - val_acc: 0.9750 - lr: 1.0000e-04 Epoch 187/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0334 - acc: 0.9908 - val_loss: 0.0334 - val_acc: 0.9862 - lr: 1.0000e-04 Epoch 188/2000 720/720 [==============================] - 3s 5ms/step - loss: 0.0332 - acc: 0.9906 - val_loss: 0.0378 - val_acc: 0.9825 - lr: 1.0000e-04 Epoch 189/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0329 - acc: 0.9904 - val_loss: 0.0425 - val_acc: 0.9800 - lr: 1.0000e-04 Epoch 190/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0328 - acc: 0.9901 - val_loss: 0.0316 - val_acc: 0.9912 - lr: 1.0000e-04 Epoch 191/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0330 - acc: 0.9911 - val_loss: 0.0463 - val_acc: 0.9800 - lr: 1.0000e-04 Epoch 192/2000 720/720 [==============================] - 2s 2ms/step - loss: 0.0326 - acc: 0.9908 - val_loss: 0.0391 - val_acc: 0.9825 - lr: 1.0000e-04
fig, ax = plt.subplots(ncols=2, figsize=(16, 6))
ax[0].plot(history.history['loss'], label='Training loss')
ax[0].plot(history.history['val_loss'], label='Validation loss')
ax[0].legend()
ax[0].set(
ylabel='Loss',
xlabel='Epoch'
)
ax[1].plot(history.history['acc'], label='Training accuracy')
ax[1].plot(history.history['val_acc'], label='Validation accuracy')
ax[1].set(
ylabel='Accuracy',
xlabel='Epoch'
)
plt.legend()
plt.tight_layout()
plt.show()
predictions = [np.argmax(p) for p in mlp.predict(x_test)]
63/63 [==============================] - 0s 1ms/step
print(classification_report(y_test, predictions))
precision recall f1-score support
0 0.99 0.98 0.98 884
1 0.98 0.99 0.98 279
2 1.00 0.98 0.99 279
3 0.98 1.00 0.99 279
4 0.99 1.00 0.99 279
accuracy 0.99 2000
macro avg 0.99 0.99 0.99 2000
weighted avg 0.99 0.99 0.99 2000
fig, ax = plt.subplots(figsize=(14, 8))
ConfusionMatrixDisplay(confusion_matrix(predictions, y_test)).plot(values_format='.0f', ax=ax)
ax.set(
title='Confusion Matrix',
xlabel='True Labels',
ylabel='Predicted Labels'
)
plt.tight_layout()
plt.show()
fig, ax = plt.subplots(figsize=(12, 12))
ax.scatter(x_test[:,0], x_test[:,1], c=y_test, marker='d')
ax.scatter(x_test[:,0], x_test[:,1], c=predictions, marker='x')
ax.set(
title='Predictions',
xlabel='$x$',
ylabel='$y$'
)
plt.show()
Considere uma rede deep learning convolutiva (treinada) aplicada à classificação de padrões em imagens. A base de dados considerada é a CIFAR-10 (pesquise). A referida base de dados consiste de 60 mil imagens coloridas de 32x32 pixels, com 50 mil para treino e 10 mil para teste. As imagens estão divididas em 10 classes, a saber: avião, navio, cami- nhão, automóvel, sapo, pássaro, cachorro, gato, cavalo e cervo. Cada imagem possui apenas um dos objetos da classe de interesse, podendo estar parcialmente obstruído por outros ob- jetos que não pertençam a esse conjunto. Apresente os resultados da classificação em uma matriz de confusão.
Este trabalho foi baseado no trabalho do CODERONIN1 do Kaggle Link: https://www.kaggle.com/code/adi160/cifar-10-keras-transfer-learning
from sklearn.utils.multiclass import unique_labels
import os
import matplotlib.image as mpimg
import seaborn as sns
from keras.layers import Flatten,Dense,BatchNormalization,Activation,Dropout
from keras.utils import to_categorical
from tensorflow.keras.callbacks import EarlyStopping
# library for transfer learning
from keras.applications import VGG19,ResNet50
# data augumentation
from keras.preprocessing.image import ImageDataGenerator
# import dataset
from keras.datasets import cifar10
# load dataset and divide in train and teste datasets
(x_train, y_train), (x_test, y_test) = cifar10.load_data()
print(f"shape x train: {x_train.shape}")
print(f"shape y train: {y_train.shape}")
print(f"shape x test: {x_test.shape}")
print(f"shape y test: {y_test.shape}")
shape x train: (50000, 32, 32, 3) shape y train: (50000, 1) shape x test: (10000, 32, 32, 3) shape y test: (10000, 1)
y_train=to_categorical(y_train)
y_test=to_categorical(y_test)
print(f"shape x train: {x_train.shape}")
print(f"shape y train: {y_train.shape}")
print(f"shape x test: {x_test.shape}")
print(f"shape y test: {y_test.shape}")
shape x train: (50000, 32, 32, 3) shape y train: (50000, 10) shape x test: (10000, 32, 32, 3) shape y test: (10000, 10)
# Let's instantiate the object
train_generator = ImageDataGenerator(
rotation_range=2,
horizontal_flip=True,
zoom_range=.1 )
test_generator = ImageDataGenerator(
rotation_range=2,
horizontal_flip= True,
zoom_range=.1)
#Fit the augmentation method to the data
train_generator.fit(x_train)
test_generator.fit(x_test)
lrr= ReduceLROnPlateau(
monitor='val_accuracy', #Metric to be measured
factor=.01, #Factor by which learning rate will be reduced
patience=3, #No. of epochs after which if there is no improvement in the val_acc, the learning rate is reduced
min_lr=1e-5) #The minimum learning rate
# The first base model used is VGG19.
# The pretrained weights from the imagenet challenge are used
base_vgg19 = VGG19(include_top=False,
input_shape=(32,32,3),
classes=y_train.shape[1],
weights='imagenet')
base_vgg19.summary()
Model: "vgg19"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
input_4 (InputLayer) [(None, 32, 32, 3)] 0
block1_conv1 (Conv2D) (None, 32, 32, 64) 1792
block1_conv2 (Conv2D) (None, 32, 32, 64) 36928
block1_pool (MaxPooling2D) (None, 16, 16, 64) 0
block2_conv1 (Conv2D) (None, 16, 16, 128) 73856
block2_conv2 (Conv2D) (None, 16, 16, 128) 147584
block2_pool (MaxPooling2D) (None, 8, 8, 128) 0
block3_conv1 (Conv2D) (None, 8, 8, 256) 295168
block3_conv2 (Conv2D) (None, 8, 8, 256) 590080
block3_conv3 (Conv2D) (None, 8, 8, 256) 590080
block3_conv4 (Conv2D) (None, 8, 8, 256) 590080
block3_pool (MaxPooling2D) (None, 4, 4, 256) 0
block4_conv1 (Conv2D) (None, 4, 4, 512) 1180160
block4_conv2 (Conv2D) (None, 4, 4, 512) 2359808
block4_conv3 (Conv2D) (None, 4, 4, 512) 2359808
block4_conv4 (Conv2D) (None, 4, 4, 512) 2359808
block4_pool (MaxPooling2D) (None, 2, 2, 512) 0
block5_conv1 (Conv2D) (None, 2, 2, 512) 2359808
block5_conv2 (Conv2D) (None, 2, 2, 512) 2359808
block5_conv3 (Conv2D) (None, 2, 2, 512) 2359808
block5_conv4 (Conv2D) (None, 2, 2, 512) 2359808
block5_pool (MaxPooling2D) (None, 1, 1, 512) 0
=================================================================
Total params: 20,024,384
Trainable params: 20,024,384
Non-trainable params: 0
_________________________________________________________________
vgg19 = Sequential()
vgg19.add(base_vgg19)
vgg19.add(Flatten())
#Add the Dense layers along with activation and batch normalization
vgg19.add(Dense(1024,activation=('relu'),input_dim=512))
vgg19.add(Dense(512,activation=('relu')))
vgg19.add(Dense(256,activation=('relu')))
#model_1.add(Dropout(.3))
vgg19.add(Dense(128,activation=('relu')))
#model_1.add(Dropout(.2))
vgg19.add(Dense(10,activation=('softmax')))
vgg19.summary()
Model: "sequential_3"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
vgg19 (Functional) (None, 1, 1, 512) 20024384
flatten_3 (Flatten) (None, 512) 0
dense_15 (Dense) (None, 1024) 525312
dense_16 (Dense) (None, 512) 524800
dense_17 (Dense) (None, 256) 131328
dense_18 (Dense) (None, 128) 32896
dense_19 (Dense) (None, 10) 1290
=================================================================
Total params: 21,240,010
Trainable params: 21,240,010
Non-trainable params: 0
_________________________________________________________________
# hyperparameters
batch_size= 100
epochs=50
learn_rate=.001
sgd=SGD(lr=learn_rate,momentum=.9,nesterov=False)
adam=Adam(lr=learn_rate, beta_1=0.9, beta_2=0.999,
epsilon=None, decay=0.0, amsgrad=False)
vgg19.compile(optimizer=sgd,
loss='categorical_crossentropy',
metrics=['accuracy'])
vgg19.fit_generator(train_generator.flow(x_train,y_train,batch_size=batch_size),
epochs=epochs,
steps_per_epoch=x_train.shape[0]//batch_size,
validation_data=test_generator.flow(x_test,y_test,batch_size=batch_size),validation_steps=250,
callbacks=[lrr],verbose=1)
Epoch 1/50
/usr/local/lib/python3.7/dist-packages/ipykernel_launcher.py:5: UserWarning: `Model.fit_generator` is deprecated and will be removed in a future version. Please use `Model.fit`, which supports generators. """
500/500 [==============================] - ETA: 0s - loss: 1.6137 - accuracy: 0.4008
WARNING:tensorflow:Your input ran out of data; interrupting training. Make sure that your dataset or generator can generate at least `steps_per_epoch * epochs` batches (in this case, 250 batches). You may need to use the repeat() function when building your dataset.
500/500 [==============================] - 37s 73ms/step - loss: 1.6137 - accuracy: 0.4008 - val_loss: 1.0790 - val_accuracy: 0.6240 - lr: 0.0010 Epoch 2/50 500/500 [==============================] - ETA: 0s - loss: 0.8359 - accuracy: 0.7123
WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,accuracy,lr
500/500 [==============================] - 32s 64ms/step - loss: 0.8359 - accuracy: 0.7123 - lr: 0.0010 Epoch 3/50 500/500 [==============================] - ETA: 0s - loss: 0.6465 - accuracy: 0.7799
WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,accuracy,lr
500/500 [==============================] - 33s 65ms/step - loss: 0.6465 - accuracy: 0.7799 - lr: 0.0010 Epoch 4/50 500/500 [==============================] - ETA: 0s - loss: 0.5492 - accuracy: 0.8131
WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,accuracy,lr
500/500 [==============================] - 32s 64ms/step - loss: 0.5492 - accuracy: 0.8131 - lr: 0.0010 Epoch 5/50 500/500 [==============================] - ETA: 0s - loss: 0.4832 - accuracy: 0.8365
WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,accuracy,lr
500/500 [==============================] - 32s 64ms/step - loss: 0.4832 - accuracy: 0.8365 - lr: 0.0010 Epoch 6/50 500/500 [==============================] - ETA: 0s - loss: 0.4242 - accuracy: 0.8548
WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,accuracy,lr
500/500 [==============================] - 32s 64ms/step - loss: 0.4242 - accuracy: 0.8548 - lr: 0.0010 Epoch 7/50 500/500 [==============================] - ETA: 0s - loss: 0.3861 - accuracy: 0.8667
WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,accuracy,lr
500/500 [==============================] - 32s 64ms/step - loss: 0.3861 - accuracy: 0.8667 - lr: 0.0010 Epoch 8/50 500/500 [==============================] - ETA: 0s - loss: 0.3509 - accuracy: 0.8775
WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,accuracy,lr
500/500 [==============================] - 35s 70ms/step - loss: 0.3509 - accuracy: 0.8775 - lr: 0.0010 Epoch 9/50 500/500 [==============================] - ETA: 0s - loss: 0.3144 - accuracy: 0.8912
WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,accuracy,lr
500/500 [==============================] - 32s 64ms/step - loss: 0.3144 - accuracy: 0.8912 - lr: 0.0010 Epoch 10/50 500/500 [==============================] - ETA: 0s - loss: 0.2832 - accuracy: 0.9022
WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,accuracy,lr
500/500 [==============================] - 32s 64ms/step - loss: 0.2832 - accuracy: 0.9022 - lr: 0.0010 Epoch 11/50 500/500 [==============================] - ETA: 0s - loss: 0.2564 - accuracy: 0.9116
WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,accuracy,lr
500/500 [==============================] - 32s 65ms/step - loss: 0.2564 - accuracy: 0.9116 - lr: 0.0010 Epoch 12/50 500/500 [==============================] - ETA: 0s - loss: 0.2257 - accuracy: 0.9214
WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,accuracy,lr
500/500 [==============================] - 32s 65ms/step - loss: 0.2257 - accuracy: 0.9214 - lr: 0.0010 Epoch 13/50 500/500 [==============================] - ETA: 0s - loss: 0.2134 - accuracy: 0.9262
WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,accuracy,lr
500/500 [==============================] - 32s 64ms/step - loss: 0.2134 - accuracy: 0.9262 - lr: 0.0010 Epoch 14/50 500/500 [==============================] - ETA: 0s - loss: 0.1928 - accuracy: 0.9343
WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,accuracy,lr
500/500 [==============================] - 32s 65ms/step - loss: 0.1928 - accuracy: 0.9343 - lr: 0.0010 Epoch 15/50 500/500 [==============================] - ETA: 0s - loss: 0.1731 - accuracy: 0.9400
WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,accuracy,lr
500/500 [==============================] - 32s 64ms/step - loss: 0.1731 - accuracy: 0.9400 - lr: 0.0010 Epoch 16/50 500/500 [==============================] - ETA: 0s - loss: 0.1550 - accuracy: 0.9474
WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,accuracy,lr
500/500 [==============================] - 32s 64ms/step - loss: 0.1550 - accuracy: 0.9474 - lr: 0.0010 Epoch 17/50 500/500 [==============================] - ETA: 0s - loss: 0.1412 - accuracy: 0.9514
WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,accuracy,lr
500/500 [==============================] - 32s 65ms/step - loss: 0.1412 - accuracy: 0.9514 - lr: 0.0010 Epoch 18/50 500/500 [==============================] - ETA: 0s - loss: 0.1307 - accuracy: 0.9560
WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,accuracy,lr
500/500 [==============================] - 32s 64ms/step - loss: 0.1307 - accuracy: 0.9560 - lr: 0.0010 Epoch 19/50 500/500 [==============================] - ETA: 0s - loss: 0.1171 - accuracy: 0.9605
WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,accuracy,lr
500/500 [==============================] - 32s 65ms/step - loss: 0.1171 - accuracy: 0.9605 - lr: 0.0010 Epoch 20/50 500/500 [==============================] - ETA: 0s - loss: 0.1071 - accuracy: 0.9633
WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,accuracy,lr
500/500 [==============================] - 32s 64ms/step - loss: 0.1071 - accuracy: 0.9633 - lr: 0.0010 Epoch 21/50 500/500 [==============================] - ETA: 0s - loss: 0.1004 - accuracy: 0.9654
WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,accuracy,lr
500/500 [==============================] - 32s 65ms/step - loss: 0.1004 - accuracy: 0.9654 - lr: 0.0010 Epoch 22/50 500/500 [==============================] - ETA: 0s - loss: 0.0919 - accuracy: 0.9687
WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,accuracy,lr
500/500 [==============================] - 32s 65ms/step - loss: 0.0919 - accuracy: 0.9687 - lr: 0.0010 Epoch 23/50 500/500 [==============================] - ETA: 0s - loss: 0.0862 - accuracy: 0.9704
WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,accuracy,lr
500/500 [==============================] - 34s 68ms/step - loss: 0.0862 - accuracy: 0.9704 - lr: 0.0010 Epoch 24/50 500/500 [==============================] - ETA: 0s - loss: 0.0761 - accuracy: 0.9743
WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,accuracy,lr
500/500 [==============================] - 32s 65ms/step - loss: 0.0761 - accuracy: 0.9743 - lr: 0.0010 Epoch 25/50 500/500 [==============================] - ETA: 0s - loss: 0.0703 - accuracy: 0.9761
WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,accuracy,lr
500/500 [==============================] - 32s 64ms/step - loss: 0.0703 - accuracy: 0.9761 - lr: 0.0010 Epoch 26/50 500/500 [==============================] - ETA: 0s - loss: 0.0664 - accuracy: 0.9768
WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,accuracy,lr
500/500 [==============================] - 32s 64ms/step - loss: 0.0664 - accuracy: 0.9768 - lr: 0.0010 Epoch 27/50 500/500 [==============================] - ETA: 0s - loss: 0.0601 - accuracy: 0.9798
WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,accuracy,lr
500/500 [==============================] - 32s 65ms/step - loss: 0.0601 - accuracy: 0.9798 - lr: 0.0010 Epoch 28/50 500/500 [==============================] - ETA: 0s - loss: 0.0578 - accuracy: 0.9802
WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,accuracy,lr
500/500 [==============================] - 32s 64ms/step - loss: 0.0578 - accuracy: 0.9802 - lr: 0.0010 Epoch 29/50 500/500 [==============================] - ETA: 0s - loss: 0.0568 - accuracy: 0.9805
WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,accuracy,lr
500/500 [==============================] - 32s 64ms/step - loss: 0.0568 - accuracy: 0.9805 - lr: 0.0010 Epoch 30/50 500/500 [==============================] - ETA: 0s - loss: 0.0488 - accuracy: 0.9842
WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,accuracy,lr
500/500 [==============================] - 32s 65ms/step - loss: 0.0488 - accuracy: 0.9842 - lr: 0.0010 Epoch 31/50 500/500 [==============================] - ETA: 0s - loss: 0.0475 - accuracy: 0.9842
WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,accuracy,lr
500/500 [==============================] - 32s 64ms/step - loss: 0.0475 - accuracy: 0.9842 - lr: 0.0010 Epoch 32/50 500/500 [==============================] - ETA: 0s - loss: 0.0455 - accuracy: 0.9844
WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,accuracy,lr
500/500 [==============================] - 32s 64ms/step - loss: 0.0455 - accuracy: 0.9844 - lr: 0.0010 Epoch 33/50 500/500 [==============================] - ETA: 0s - loss: 0.0433 - accuracy: 0.9855
WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,accuracy,lr
500/500 [==============================] - 32s 65ms/step - loss: 0.0433 - accuracy: 0.9855 - lr: 0.0010 Epoch 34/50 500/500 [==============================] - ETA: 0s - loss: 0.0404 - accuracy: 0.9866
WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,accuracy,lr
500/500 [==============================] - 32s 65ms/step - loss: 0.0404 - accuracy: 0.9866 - lr: 0.0010 Epoch 35/50 500/500 [==============================] - ETA: 0s - loss: 0.0433 - accuracy: 0.9849
WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,accuracy,lr
500/500 [==============================] - 32s 64ms/step - loss: 0.0433 - accuracy: 0.9849 - lr: 0.0010 Epoch 36/50 500/500 [==============================] - ETA: 0s - loss: 0.0367 - accuracy: 0.9876
WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,accuracy,lr
500/500 [==============================] - 32s 65ms/step - loss: 0.0367 - accuracy: 0.9876 - lr: 0.0010 Epoch 37/50 500/500 [==============================] - ETA: 0s - loss: 0.0306 - accuracy: 0.9895
WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,accuracy,lr
500/500 [==============================] - 32s 64ms/step - loss: 0.0306 - accuracy: 0.9895 - lr: 0.0010 Epoch 38/50 500/500 [==============================] - ETA: 0s - loss: 0.0300 - accuracy: 0.9896
WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,accuracy,lr
500/500 [==============================] - 32s 65ms/step - loss: 0.0300 - accuracy: 0.9896 - lr: 0.0010 Epoch 39/50 500/500 [==============================] - ETA: 0s - loss: 0.0328 - accuracy: 0.9892
WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,accuracy,lr
500/500 [==============================] - 32s 65ms/step - loss: 0.0328 - accuracy: 0.9892 - lr: 0.0010 Epoch 40/50 500/500 [==============================] - ETA: 0s - loss: 0.0263 - accuracy: 0.9912
WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,accuracy,lr
500/500 [==============================] - 32s 65ms/step - loss: 0.0263 - accuracy: 0.9912 - lr: 0.0010 Epoch 41/50 500/500 [==============================] - ETA: 0s - loss: 0.0276 - accuracy: 0.9909
WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,accuracy,lr
500/500 [==============================] - 32s 64ms/step - loss: 0.0276 - accuracy: 0.9909 - lr: 0.0010 Epoch 42/50 500/500 [==============================] - ETA: 0s - loss: 0.0274 - accuracy: 0.9908
WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,accuracy,lr
500/500 [==============================] - 33s 65ms/step - loss: 0.0274 - accuracy: 0.9908 - lr: 0.0010 Epoch 43/50 500/500 [==============================] - ETA: 0s - loss: 0.0267 - accuracy: 0.9907
WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,accuracy,lr
500/500 [==============================] - 32s 65ms/step - loss: 0.0267 - accuracy: 0.9907 - lr: 0.0010 Epoch 44/50 500/500 [==============================] - ETA: 0s - loss: 0.0308 - accuracy: 0.9895
WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,accuracy,lr
500/500 [==============================] - 32s 65ms/step - loss: 0.0308 - accuracy: 0.9895 - lr: 0.0010 Epoch 45/50 500/500 [==============================] - ETA: 0s - loss: 0.0257 - accuracy: 0.9914
WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,accuracy,lr
500/500 [==============================] - 32s 64ms/step - loss: 0.0257 - accuracy: 0.9914 - lr: 0.0010 Epoch 46/50 500/500 [==============================] - ETA: 0s - loss: 0.0220 - accuracy: 0.9930
WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,accuracy,lr
500/500 [==============================] - 32s 64ms/step - loss: 0.0220 - accuracy: 0.9930 - lr: 0.0010 Epoch 47/50 500/500 [==============================] - ETA: 0s - loss: 0.0216 - accuracy: 0.9928
WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,accuracy,lr
500/500 [==============================] - 32s 65ms/step - loss: 0.0216 - accuracy: 0.9928 - lr: 0.0010 Epoch 48/50 500/500 [==============================] - ETA: 0s - loss: 0.0211 - accuracy: 0.9928
WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,accuracy,lr
500/500 [==============================] - 32s 64ms/step - loss: 0.0211 - accuracy: 0.9928 - lr: 0.0010 Epoch 49/50 500/500 [==============================] - ETA: 0s - loss: 0.0216 - accuracy: 0.9926
WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,accuracy,lr
500/500 [==============================] - 33s 65ms/step - loss: 0.0216 - accuracy: 0.9926 - lr: 0.0010 Epoch 50/50 500/500 [==============================] - ETA: 0s - loss: 0.0230 - accuracy: 0.9920
WARNING:tensorflow:Learning rate reduction is conditioned on metric `val_accuracy` which is not available. Available metrics are: loss,accuracy,lr
500/500 [==============================] - 32s 64ms/step - loss: 0.0230 - accuracy: 0.9920 - lr: 0.0010
<keras.callbacks.History at 0x7f660232e310>
fig, ax = plt.subplots(ncols=2, figsize=(16, 6))
ax[0].plot(vgg19.history.history['loss'], label='Training loss')
ax[0].plot(vgg19.history.history['val_loss'], label='Validation loss')
ax[0].legend()
ax[0].set(
ylabel='Loss',
xlabel='Epoch'
)
ax[1].plot(vgg19.history.history['accuracy'], label='Training accuracy')
ax[1].plot(vgg19.history.history['val_accuracy'], label='Validation accuracy')
ax[1].set(
ylabel='Accuracy',
xlabel='Epoch'
)
plt.legend()
plt.tight_layout()
plt.show()
def plot_confusion_matrix(y_true, y_pred, classes,
normalize=False,
title=None,
cmap=plt.cm.Blues):
"""
This function prints and plots the confusion matrix.
Normalization can be applied by setting `normalize=True`.
"""
if not title:
if normalize:
title = 'Normalized confusion matrix'
else:
title = 'Confusion matrix, without normalization'
# Compute confusion matrix
cm = confusion_matrix(y_true, y_pred)
if normalize:
cm = cm.astype('float') / cm.sum(axis=1)[:, np.newaxis]
print("Normalized confusion matrix")
else:
print('Confusion matrix, without normalization')
# print(cm)
fig, ax = plt.subplots(figsize=(7,7))
im = ax.imshow(cm, interpolation='nearest', cmap=cmap)
ax.figure.colorbar(im, ax=ax)
# We want to show all ticks...
ax.set(xticks=np.arange(cm.shape[1]),
yticks=np.arange(cm.shape[0]),
# ... and label them with the respective list entries
xticklabels=classes, yticklabels=classes,
title=title,
ylabel='True label',
xlabel='Predicted label')
# Rotate the tick labels and set their alignment.
plt.setp(ax.get_xticklabels(), rotation=45, ha="right",
rotation_mode="anchor")
# Loop over data dimensions and create text annotations.
fmt = '.2f' if normalize else 'd'
thresh = cm.max() / 2.
for i in range(cm.shape[0]):
for j in range(cm.shape[1]):
ax.text(j, i, format(cm[i, j], fmt),
ha="center", va="center",
color="white" if cm[i, j] > thresh else "black")
fig.tight_layout()
return ax
np.set_printoptions(precision=2)
y_pred = vgg19.predict(x_test)
y_pred = np.argmax(y_pred,axis=1)
y_true=np.argmax(y_test,axis=1)
#Compute the confusion matrix
confusion_mtx = confusion_matrix(y_true,y_pred)
313/313 [==============================] - 4s 10ms/step
class_names=['airplane', 'automobile', 'bird', 'cat', 'deer', 'dog', 'frog', 'horse', 'ship', 'truck']
# Plot non-normalized confusion matrix
plot_confusion_matrix(y_true, y_pred, classes=class_names,
title='Confusion matrix, without normalization')
Confusion matrix, without normalization
<matplotlib.axes._subplots.AxesSubplot at 0x7f64c4167a10>
# Plot normalized confusion matrix
plot_confusion_matrix(y_true, y_pred, classes=class_names, normalize=True,
title='Normalized confusion matrix')
# plt.show()
Normalized confusion matrix
<matplotlib.axes._subplots.AxesSubplot at 0x7f64ba2477d0>
print(classification_report(y_true, y_pred, target_names=class_names))
precision recall f1-score support
airplane 0.89 0.89 0.89 1000
automobile 0.91 0.95 0.93 1000
bird 0.90 0.80 0.85 1000
cat 0.77 0.70 0.74 1000
deer 0.86 0.87 0.86 1000
dog 0.78 0.82 0.80 1000
frog 0.89 0.90 0.90 1000
horse 0.89 0.90 0.89 1000
ship 0.93 0.93 0.93 1000
truck 0.87 0.94 0.90 1000
accuracy 0.87 10000
macro avg 0.87 0.87 0.87 10000
weighted avg 0.87 0.87 0.87 10000
Utilize a rede neural perceptron de múltiplas camadas do tipo NARX (rede recorrente) para fazer a predição de um passo $x^{(n+1)}$ da série temporal $x(n) = 1 + cos(n + cos^2(n))$, n=0,1,2,3,.... Gere inicialmente um conjunto de amostras para o treinamento, definindo o erro de predição como $e^{(n+1)}=x(n+1)-x^{(n+1)}$. Avalie o desempenho mostrando a curva a série temporal, a curva de predição o e curva do erro de predição.
from tensorflow.keras.layers import Dense, Reshape, Flatten, Dropout, Activation, BatchNormalization, LSTM, Embedding, Input
from tensorflow.keras.preprocessing.sequence import TimeseriesGenerator
from tensorflow.keras.callbacks import EarlyStopping, ReduceLROnPlateau
# defining function
def f(x):
return 1 + np.cos(x + (np.cos(x))**2)
x = np.linspace(0, 100, 10000)
y = [f(i) for i in x]
# plot data
points = 10000
plt.plot(x[:points], y[:points])
plt.show()
# splitting data into training and testing
test_size = 2000
x_train = x[:-test_size]
y_train = y[:-test_size]
x_test = x[-test_size:]
y_test = y[-test_size:]
# plot data
fig, axes = plt.subplots(ncols=2, figsize=(24, 5))
axes[0].plot(x_train, y_train, label='train')
axes[0].plot(x_test, y_test, label='test')
axes[0].set_title('Data')
axes[0].legend()
axes[1].plot(x_train, y_train, label='train')
axes[1].plot(x_test, y_test, label='test')
axes[1].set_title('Data')
axes[1].legend()
axes[1].set_xlim(0, 100)
axes[1].set_ylim(0, 2)
plt.show()
sequence_x = list(TimeseriesGenerator(x_train, x_train, 4, batch_size=1))
sequence_y = list(TimeseriesGenerator(x_train[4:], x_train[4:], 3, batch_size=1))
train_seqs = []
y_train = []
for (x_seq, next_x), (next_seq, _) in zip(sequence_x, sequence_y):
seq = np.append(x_seq.reshape(4,), next_seq.reshape(3, ))
train_seqs.append(seq)
y_train.append(next_x)
train_seqs = np.array(train_seqs)
y_train = np.array(y_train)
sequence_x = list(TimeseriesGenerator(x_test, x_test, 4, batch_size=1))
sequence_y = list(TimeseriesGenerator(x_test[4:], x_test[4:], 3, batch_size=1))
test_seqs = []
y_test = []
for (x_seq, next_x), (next_seq, _) in zip(sequence_x, sequence_y):
seq = np.append(x_seq.reshape(4,), next_seq.reshape(3, ))
test_seqs.append(seq)
y_test.append(next_x)
test_seqs = np.array(test_seqs)
y_test = np.array(y_test)
# building the model
model = Sequential([
LSTM(128, input_shape=(7, 1), return_sequences=True),
LSTM(64),
Dense(1)
])
model.compile(loss="mean_squared_error", optimizer="adam")
model.summary()
Model: "sequential"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
lstm (LSTM) (None, 7, 128) 66560
lstm_1 (LSTM) (None, 64) 49408
dense (Dense) (None, 1) 65
=================================================================
Total params: 116,033
Trainable params: 116,033
Non-trainable params: 0
_________________________________________________________________
# Training the model
history = model.fit(
train_seqs, y_train,
validation_split=0.1,
batch_size=8,
epochs=100,
shuffle=True,
callbacks=[
EarlyStopping(monitor='val_loss', patience=5),
ReduceLROnPlateau(monitor='val_loss', factor=0.1, patience=5, min_lr=0.0001)
]
)
Epoch 1/100 900/900 [==============================] - 18s 15ms/step - loss: 396.9484 - val_loss: 888.2201 - lr: 0.0010 Epoch 2/100 900/900 [==============================] - 12s 14ms/step - loss: 30.8379 - val_loss: 269.7425 - lr: 0.0010 Epoch 3/100 900/900 [==============================] - 12s 14ms/step - loss: 3.9125 - val_loss: 116.8918 - lr: 0.0010 Epoch 4/100 900/900 [==============================] - 12s 14ms/step - loss: 0.6845 - val_loss: 65.4554 - lr: 0.0010 Epoch 5/100 900/900 [==============================] - 12s 13ms/step - loss: 0.3236 - val_loss: 44.9251 - lr: 0.0010 Epoch 6/100 900/900 [==============================] - 12s 14ms/step - loss: 0.0635 - val_loss: 34.4474 - lr: 0.0010 Epoch 7/100 900/900 [==============================] - 12s 13ms/step - loss: 0.1411 - val_loss: 29.1309 - lr: 0.0010 Epoch 8/100 900/900 [==============================] - 12s 13ms/step - loss: 0.0852 - val_loss: 26.6374 - lr: 0.0010 Epoch 9/100 900/900 [==============================] - 13s 15ms/step - loss: 0.0464 - val_loss: 24.4883 - lr: 0.0010 Epoch 10/100 900/900 [==============================] - 12s 13ms/step - loss: 0.0964 - val_loss: 23.4697 - lr: 0.0010 Epoch 11/100 900/900 [==============================] - 12s 13ms/step - loss: 0.0574 - val_loss: 23.2975 - lr: 0.0010 Epoch 12/100 900/900 [==============================] - 12s 14ms/step - loss: 0.0649 - val_loss: 54.3588 - lr: 0.0010 Epoch 13/100 900/900 [==============================] - 12s 14ms/step - loss: 0.0961 - val_loss: 25.9568 - lr: 0.0010 Epoch 14/100 900/900 [==============================] - 12s 13ms/step - loss: 0.0266 - val_loss: 22.9685 - lr: 0.0010 Epoch 15/100 900/900 [==============================] - 13s 14ms/step - loss: 0.0913 - val_loss: 23.4417 - lr: 0.0010 Epoch 16/100 900/900 [==============================] - 13s 15ms/step - loss: 0.0476 - val_loss: 23.7311 - lr: 0.0010 Epoch 17/100 900/900 [==============================] - 13s 15ms/step - loss: 0.0251 - val_loss: 21.4442 - lr: 0.0010 Epoch 18/100 900/900 [==============================] - 13s 15ms/step - loss: 0.0392 - val_loss: 20.8180 - lr: 0.0010 Epoch 19/100 900/900 [==============================] - 13s 14ms/step - loss: 0.0680 - val_loss: 19.8773 - lr: 0.0010 Epoch 20/100 900/900 [==============================] - 12s 14ms/step - loss: 0.0369 - val_loss: 21.5628 - lr: 0.0010 Epoch 21/100 900/900 [==============================] - 12s 14ms/step - loss: 0.0423 - val_loss: 22.3235 - lr: 0.0010 Epoch 22/100 900/900 [==============================] - 12s 13ms/step - loss: 0.0675 - val_loss: 18.6382 - lr: 0.0010 Epoch 23/100 900/900 [==============================] - 13s 15ms/step - loss: 0.0404 - val_loss: 22.6039 - lr: 0.0010 Epoch 24/100 900/900 [==============================] - 13s 14ms/step - loss: 0.0613 - val_loss: 18.8377 - lr: 0.0010 Epoch 25/100 900/900 [==============================] - 13s 14ms/step - loss: 0.0054 - val_loss: 18.0627 - lr: 0.0010 Epoch 26/100 900/900 [==============================] - 13s 14ms/step - loss: 0.0381 - val_loss: 19.9228 - lr: 0.0010 Epoch 27/100 900/900 [==============================] - 13s 15ms/step - loss: 0.0430 - val_loss: 21.3316 - lr: 0.0010 Epoch 28/100 900/900 [==============================] - 12s 13ms/step - loss: 0.0353 - val_loss: 19.2926 - lr: 0.0010 Epoch 29/100 900/900 [==============================] - 12s 14ms/step - loss: 0.0191 - val_loss: 17.4730 - lr: 0.0010 Epoch 30/100 900/900 [==============================] - 13s 15ms/step - loss: 0.0306 - val_loss: 17.8915 - lr: 0.0010 Epoch 31/100 900/900 [==============================] - 12s 14ms/step - loss: 0.0687 - val_loss: 19.3206 - lr: 0.0010 Epoch 32/100 900/900 [==============================] - 12s 14ms/step - loss: 0.0118 - val_loss: 18.8057 - lr: 0.0010 Epoch 33/100 900/900 [==============================] - 12s 13ms/step - loss: 0.0397 - val_loss: 22.1194 - lr: 0.0010 Epoch 34/100 900/900 [==============================] - 12s 13ms/step - loss: 0.0425 - val_loss: 19.9963 - lr: 0.0010
# plotting loss
fig, ax = plt.subplots(figsize=(8, 6))
ax.plot(history.history['loss'], label='Training loss')
ax.plot(history.history['val_loss'], label='Validation loss')
ax.legend()
ax.set(
ylabel='Loss',
xlabel='Epoch'
)
plt.legend()
plt.tight_layout()
plt.show()
# predicting
y_pred = model.predict(test_seqs)
63/63 [==============================] - 3s 12ms/step
fig, axes = plt.subplots(ncols=2, figsize=(25, 6))
axes[0].plot(y_test, marker='.', label='Groundtruth')
axes[0].scatter(
range(len(y_pred)), y_pred,
marker='X', edgecolors='k',
label='Predictions', c='#ff7f0e',
s=64
)
axes[0].legend()
axes[0].set(
ylabel='$\hat{x}(n)$',
xlabel='$n$'
)
axes[1].plot(y_test[:50], marker='.', label='Groundtruth')
axes[1].scatter(
range(len(y_pred[:50])), y_pred[:50],
marker='X', edgecolors='k',
label='Predictions', c='#ff7f0e',
s=64
)
axes[1].legend()
axes[1].set(
ylabel='$\hat{x}(n)$',
xlabel='$n$'
)
plt.tight_layout()
plt.show()
Considere quatro distribuições gaussianas, $C_1, C_2, C_3, C_4$, em um espaço de entrada de dimensionalidade igual a oito, isto é x = ($x_1, x_2, ..., x_8)^t$ .Todas as nuvens de dados formadas têm variâncias unitária, mas centros ou vetores média são diferentes e dados por $m_1 = (0,0,0,0,0,0,0,0)^t , m_2 = (4,0,0,0,0,0,0,0)^t, m_3 = (0,0,0,4,0,0,0,0)^t, m_4 = (0,0,0,0,0,0,0,4)^t $.
Utilize uma rede de autoeconders para reduzir a dimensionalidade dos dados para duas di- mensões. O objetivo é visualizar os dados de dimensão 8 em um espaço de dimensão 2. Esboce os dados neste novo espaço.
Observação: Gere inicialmente os dados em dimensão oito para cada uma das distribuições gaussianas. Selecione o conjunto de treinamento. Defina uma rede de autoencoder com uma arquitetura por exemplo do tipo 8:2:8 ou outro equivalente com mais camadas mas que reduza para 2 dimensões. Após o treinamento faça a redução de dimensionalidade com a rede de arquitetura 8:2 por exemplo.
from tensorflow.keras import Model
# generate four gaussian distributions with 8 dimensions, mean m1 = (0,0,0,0,0,0,0,0), m2 = (4,0,0,0,0,0,0,0), m3 = (0,0,0,4,0,0,0,0), m4 = (0,0,0,0,0,0,0,4) and covariance matrix I
m1 = np.zeros(8)
m2 = np.array([4,0,0,0,0,0,0,0])
m3 = np.array([0,0,0,4,0,0,0,0])
m4 = np.array([0,0,0,0,0,0,0,4])
I = np.eye(8)
x1 = np.random.multivariate_normal(m1, I, 1000)
x2 = np.random.multivariate_normal(m2, I, 1000)
x3 = np.random.multivariate_normal(m3, I, 1000)
x4 = np.random.multivariate_normal(m4, I, 1000)
# plot the data
fig, axes = plt.subplots(ncols=2, figsize=(25, 6))
axes[0].scatter(x1[:,0], x1[:,1], label='Class 1')
axes[0].scatter(x2[:,0], x2[:,1], label='Class 2')
axes[0].scatter(x3[:,4], x3[:,3], label='Class 3')
axes[0].scatter(x4[:,-1], x4[:,-2], label='Class 4')
axes[0].legend()
axes[0].set(
ylabel='$x_2$',
xlabel='$x_1$'
)
plt.show()
# create an autoencoder netork to reduce the dimensionality of the data to 2 and plot the data
input_dim = Input(shape=(8,))
encoded = Dense(4, activation='relu')(input_dim)
encoded = Dense(2, activation='relu')(input_dim)
decoded = Dense(4, activation='relu')(encoded)
decoded = Dense(8, activation='linear')(encoded)
autoencoder = Model(input_dim, decoded)
encoder = Model(input_dim, encoded)
autoencoder.compile(optimizer='rmsprop', loss='mean_absolute_error')
history1 = autoencoder.fit(x1, x1, epochs=300, batch_size=256, shuffle=True, validation_split=0.1)
history2 = autoencoder.fit(x2, x2, epochs=300, batch_size=256, shuffle=True, validation_split=0.1)
history4 = autoencoder.fit(x3, x3, epochs=300, batch_size=256, shuffle=True, validation_split=0.1)
history3 = autoencoder.fit(x4, x4, epochs=300, batch_size=256, shuffle=True, validation_split=0.1)
encoded_x1 = encoder.predict(x1)
encoded_x2 = encoder.predict(x2)
encoded_x3 = encoder.predict(x3)
encoded_x4 = encoder.predict(x4)
Epoch 1/300 4/4 [==============================] - 0s 42ms/step - loss: 0.9442 - val_loss: 0.9154 Epoch 2/300 4/4 [==============================] - 0s 8ms/step - loss: 0.9352 - val_loss: 0.9069 Epoch 3/300 4/4 [==============================] - 0s 8ms/step - loss: 0.9290 - val_loss: 0.9002 Epoch 4/300 4/4 [==============================] - 0s 10ms/step - loss: 0.9239 - val_loss: 0.8942 Epoch 5/300 4/4 [==============================] - 0s 9ms/step - loss: 0.9192 - val_loss: 0.8886 Epoch 6/300 4/4 [==============================] - 0s 9ms/step - loss: 0.9148 - val_loss: 0.8832 Epoch 7/300 4/4 [==============================] - 0s 9ms/step - loss: 0.9107 - val_loss: 0.8781 Epoch 8/300 4/4 [==============================] - 0s 9ms/step - loss: 0.9066 - val_loss: 0.8731 Epoch 9/300 4/4 [==============================] - 0s 9ms/step - loss: 0.9028 - val_loss: 0.8683 Epoch 10/300 4/4 [==============================] - 0s 11ms/step - loss: 0.8990 - val_loss: 0.8638 Epoch 11/300 4/4 [==============================] - 0s 9ms/step - loss: 0.8955 - val_loss: 0.8592 Epoch 12/300 4/4 [==============================] - 0s 9ms/step - loss: 0.8920 - val_loss: 0.8549 Epoch 13/300 4/4 [==============================] - 0s 9ms/step - loss: 0.8886 - val_loss: 0.8506 Epoch 14/300 4/4 [==============================] - 0s 9ms/step - loss: 0.8853 - val_loss: 0.8465 Epoch 15/300 4/4 [==============================] - 0s 9ms/step - loss: 0.8821 - val_loss: 0.8425 Epoch 16/300 4/4 [==============================] - 0s 9ms/step - loss: 0.8790 - val_loss: 0.8383 Epoch 17/300 4/4 [==============================] - 0s 10ms/step - loss: 0.8759 - val_loss: 0.8344 Epoch 18/300 4/4 [==============================] - 0s 10ms/step - loss: 0.8730 - val_loss: 0.8307 Epoch 19/300 4/4 [==============================] - 0s 9ms/step - loss: 0.8701 - val_loss: 0.8270 Epoch 20/300 4/4 [==============================] - 0s 9ms/step - loss: 0.8672 - val_loss: 0.8235 Epoch 21/300 4/4 [==============================] - 0s 9ms/step - loss: 0.8645 - val_loss: 0.8201 Epoch 22/300 4/4 [==============================] - 0s 8ms/step - loss: 0.8617 - val_loss: 0.8167 Epoch 23/300 4/4 [==============================] - 0s 9ms/step - loss: 0.8590 - val_loss: 0.8133 Epoch 24/300 4/4 [==============================] - 0s 10ms/step - loss: 0.8564 - val_loss: 0.8104 Epoch 25/300 4/4 [==============================] - 0s 10ms/step - loss: 0.8540 - val_loss: 0.8076 Epoch 26/300 4/4 [==============================] - 0s 8ms/step - loss: 0.8516 - val_loss: 0.8045 Epoch 27/300 4/4 [==============================] - 0s 8ms/step - loss: 0.8491 - val_loss: 0.8016 Epoch 28/300 4/4 [==============================] - 0s 8ms/step - loss: 0.8468 - val_loss: 0.7988 Epoch 29/300 4/4 [==============================] - 0s 9ms/step - loss: 0.8445 - val_loss: 0.7961 Epoch 30/300 4/4 [==============================] - 0s 9ms/step - loss: 0.8423 - val_loss: 0.7935 Epoch 31/300 4/4 [==============================] - 0s 9ms/step - loss: 0.8402 - val_loss: 0.7911 Epoch 32/300 4/4 [==============================] - 0s 8ms/step - loss: 0.8381 - val_loss: 0.7887 Epoch 33/300 4/4 [==============================] - 0s 9ms/step - loss: 0.8360 - val_loss: 0.7864 Epoch 34/300 4/4 [==============================] - 0s 16ms/step - loss: 0.8340 - val_loss: 0.7842 Epoch 35/300 4/4 [==============================] - 0s 9ms/step - loss: 0.8321 - val_loss: 0.7819 Epoch 36/300 4/4 [==============================] - 0s 9ms/step - loss: 0.8302 - val_loss: 0.7799 Epoch 37/300 4/4 [==============================] - 0s 11ms/step - loss: 0.8283 - val_loss: 0.7778 Epoch 38/300 4/4 [==============================] - 0s 9ms/step - loss: 0.8265 - val_loss: 0.7758 Epoch 39/300 4/4 [==============================] - 0s 9ms/step - loss: 0.8248 - val_loss: 0.7740 Epoch 40/300 4/4 [==============================] - 0s 9ms/step - loss: 0.8231 - val_loss: 0.7721 Epoch 41/300 4/4 [==============================] - 0s 10ms/step - loss: 0.8214 - val_loss: 0.7702 Epoch 42/300 4/4 [==============================] - 0s 9ms/step - loss: 0.8197 - val_loss: 0.7684 Epoch 43/300 4/4 [==============================] - 0s 9ms/step - loss: 0.8181 - val_loss: 0.7667 Epoch 44/300 4/4 [==============================] - 0s 9ms/step - loss: 0.8165 - val_loss: 0.7650 Epoch 45/300 4/4 [==============================] - 0s 8ms/step - loss: 0.8149 - val_loss: 0.7633 Epoch 46/300 4/4 [==============================] - 0s 9ms/step - loss: 0.8135 - val_loss: 0.7616 Epoch 47/300 4/4 [==============================] - 0s 8ms/step - loss: 0.8120 - val_loss: 0.7600 Epoch 48/300 4/4 [==============================] - 0s 8ms/step - loss: 0.8105 - val_loss: 0.7583 Epoch 49/300 4/4 [==============================] - 0s 15ms/step - loss: 0.8090 - val_loss: 0.7568 Epoch 50/300 4/4 [==============================] - 0s 9ms/step - loss: 0.8076 - val_loss: 0.7553 Epoch 51/300 4/4 [==============================] - 0s 9ms/step - loss: 0.8062 - val_loss: 0.7537 Epoch 52/300 4/4 [==============================] - 0s 10ms/step - loss: 0.8048 - val_loss: 0.7523 Epoch 53/300 4/4 [==============================] - 0s 9ms/step - loss: 0.8035 - val_loss: 0.7510 Epoch 54/300 4/4 [==============================] - 0s 9ms/step - loss: 0.8022 - val_loss: 0.7497 Epoch 55/300 4/4 [==============================] - 0s 8ms/step - loss: 0.8009 - val_loss: 0.7483 Epoch 56/300 4/4 [==============================] - 0s 10ms/step - loss: 0.7996 - val_loss: 0.7470 Epoch 57/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7984 - val_loss: 0.7457 Epoch 58/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7972 - val_loss: 0.7444 Epoch 59/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7960 - val_loss: 0.7431 Epoch 60/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7948 - val_loss: 0.7419 Epoch 61/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7936 - val_loss: 0.7407 Epoch 62/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7925 - val_loss: 0.7393 Epoch 63/300 4/4 [==============================] - 0s 11ms/step - loss: 0.7914 - val_loss: 0.7381 Epoch 64/300 4/4 [==============================] - 0s 15ms/step - loss: 0.7902 - val_loss: 0.7369 Epoch 65/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7891 - val_loss: 0.7357 Epoch 66/300 4/4 [==============================] - 0s 8ms/step - loss: 0.7880 - val_loss: 0.7345 Epoch 67/300 4/4 [==============================] - 0s 10ms/step - loss: 0.7869 - val_loss: 0.7333 Epoch 68/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7858 - val_loss: 0.7321 Epoch 69/300 4/4 [==============================] - 0s 10ms/step - loss: 0.7848 - val_loss: 0.7310 Epoch 70/300 4/4 [==============================] - 0s 8ms/step - loss: 0.7837 - val_loss: 0.7300 Epoch 71/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7827 - val_loss: 0.7289 Epoch 72/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7816 - val_loss: 0.7278 Epoch 73/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7805 - val_loss: 0.7267 Epoch 74/300 4/4 [==============================] - 0s 8ms/step - loss: 0.7795 - val_loss: 0.7257 Epoch 75/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7784 - val_loss: 0.7246 Epoch 76/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7774 - val_loss: 0.7235 Epoch 77/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7763 - val_loss: 0.7224 Epoch 78/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7753 - val_loss: 0.7214 Epoch 79/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7743 - val_loss: 0.7203 Epoch 80/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7733 - val_loss: 0.7193 Epoch 81/300 4/4 [==============================] - 0s 10ms/step - loss: 0.7723 - val_loss: 0.7184 Epoch 82/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7713 - val_loss: 0.7175 Epoch 83/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7703 - val_loss: 0.7165 Epoch 84/300 4/4 [==============================] - 0s 11ms/step - loss: 0.7693 - val_loss: 0.7157 Epoch 85/300 4/4 [==============================] - 0s 11ms/step - loss: 0.7683 - val_loss: 0.7148 Epoch 86/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7673 - val_loss: 0.7139 Epoch 87/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7663 - val_loss: 0.7131 Epoch 88/300 4/4 [==============================] - 0s 8ms/step - loss: 0.7653 - val_loss: 0.7122 Epoch 89/300 4/4 [==============================] - 0s 11ms/step - loss: 0.7644 - val_loss: 0.7112 Epoch 90/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7634 - val_loss: 0.7104 Epoch 91/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7625 - val_loss: 0.7096 Epoch 92/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7615 - val_loss: 0.7087 Epoch 93/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7606 - val_loss: 0.7079 Epoch 94/300 4/4 [==============================] - 0s 8ms/step - loss: 0.7597 - val_loss: 0.7070 Epoch 95/300 4/4 [==============================] - 0s 11ms/step - loss: 0.7588 - val_loss: 0.7061 Epoch 96/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7579 - val_loss: 0.7052 Epoch 97/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7571 - val_loss: 0.7043 Epoch 98/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7562 - val_loss: 0.7034 Epoch 99/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7553 - val_loss: 0.7025 Epoch 100/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7545 - val_loss: 0.7018 Epoch 101/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7537 - val_loss: 0.7010 Epoch 102/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7528 - val_loss: 0.7001 Epoch 103/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7520 - val_loss: 0.6993 Epoch 104/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7512 - val_loss: 0.6984 Epoch 105/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7503 - val_loss: 0.6976 Epoch 106/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7496 - val_loss: 0.6968 Epoch 107/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7488 - val_loss: 0.6960 Epoch 108/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7480 - val_loss: 0.6952 Epoch 109/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7473 - val_loss: 0.6944 Epoch 110/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7465 - val_loss: 0.6936 Epoch 111/300 4/4 [==============================] - 0s 8ms/step - loss: 0.7458 - val_loss: 0.6928 Epoch 112/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7451 - val_loss: 0.6921 Epoch 113/300 4/4 [==============================] - 0s 15ms/step - loss: 0.7444 - val_loss: 0.6915 Epoch 114/300 4/4 [==============================] - 0s 8ms/step - loss: 0.7437 - val_loss: 0.6907 Epoch 115/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7430 - val_loss: 0.6900 Epoch 116/300 4/4 [==============================] - 0s 16ms/step - loss: 0.7423 - val_loss: 0.6894 Epoch 117/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7417 - val_loss: 0.6887 Epoch 118/300 4/4 [==============================] - 0s 10ms/step - loss: 0.7411 - val_loss: 0.6880 Epoch 119/300 4/4 [==============================] - 0s 10ms/step - loss: 0.7404 - val_loss: 0.6873 Epoch 120/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7398 - val_loss: 0.6867 Epoch 121/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7391 - val_loss: 0.6861 Epoch 122/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7385 - val_loss: 0.6855 Epoch 123/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7380 - val_loss: 0.6848 Epoch 124/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7374 - val_loss: 0.6842 Epoch 125/300 4/4 [==============================] - 0s 10ms/step - loss: 0.7368 - val_loss: 0.6836 Epoch 126/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7362 - val_loss: 0.6829 Epoch 127/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7356 - val_loss: 0.6824 Epoch 128/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7351 - val_loss: 0.6818 Epoch 129/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7345 - val_loss: 0.6813 Epoch 130/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7340 - val_loss: 0.6806 Epoch 131/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7334 - val_loss: 0.6801 Epoch 132/300 4/4 [==============================] - 0s 8ms/step - loss: 0.7329 - val_loss: 0.6796 Epoch 133/300 4/4 [==============================] - 0s 8ms/step - loss: 0.7324 - val_loss: 0.6791 Epoch 134/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7319 - val_loss: 0.6786 Epoch 135/300 4/4 [==============================] - 0s 10ms/step - loss: 0.7314 - val_loss: 0.6781 Epoch 136/300 4/4 [==============================] - 0s 8ms/step - loss: 0.7309 - val_loss: 0.6777 Epoch 137/300 4/4 [==============================] - 0s 10ms/step - loss: 0.7304 - val_loss: 0.6772 Epoch 138/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7299 - val_loss: 0.6768 Epoch 139/300 4/4 [==============================] - 0s 8ms/step - loss: 0.7294 - val_loss: 0.6764 Epoch 140/300 4/4 [==============================] - 0s 8ms/step - loss: 0.7289 - val_loss: 0.6759 Epoch 141/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7284 - val_loss: 0.6754 Epoch 142/300 4/4 [==============================] - 0s 11ms/step - loss: 0.7279 - val_loss: 0.6751 Epoch 143/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7275 - val_loss: 0.6747 Epoch 144/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7271 - val_loss: 0.6743 Epoch 145/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7266 - val_loss: 0.6739 Epoch 146/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7262 - val_loss: 0.6735 Epoch 147/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7258 - val_loss: 0.6731 Epoch 148/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7254 - val_loss: 0.6726 Epoch 149/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7250 - val_loss: 0.6721 Epoch 150/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7246 - val_loss: 0.6717 Epoch 151/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7242 - val_loss: 0.6713 Epoch 152/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7238 - val_loss: 0.6710 Epoch 153/300 4/4 [==============================] - 0s 10ms/step - loss: 0.7234 - val_loss: 0.6706 Epoch 154/300 4/4 [==============================] - 0s 10ms/step - loss: 0.7230 - val_loss: 0.6702 Epoch 155/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7226 - val_loss: 0.6698 Epoch 156/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7222 - val_loss: 0.6695 Epoch 157/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7219 - val_loss: 0.6691 Epoch 158/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7215 - val_loss: 0.6686 Epoch 159/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7212 - val_loss: 0.6683 Epoch 160/300 4/4 [==============================] - 0s 8ms/step - loss: 0.7208 - val_loss: 0.6679 Epoch 161/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7205 - val_loss: 0.6673 Epoch 162/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7201 - val_loss: 0.6669 Epoch 163/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7198 - val_loss: 0.6666 Epoch 164/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7195 - val_loss: 0.6662 Epoch 165/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7191 - val_loss: 0.6659 Epoch 166/300 4/4 [==============================] - 0s 10ms/step - loss: 0.7188 - val_loss: 0.6656 Epoch 167/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7185 - val_loss: 0.6651 Epoch 168/300 4/4 [==============================] - 0s 13ms/step - loss: 0.7182 - val_loss: 0.6648 Epoch 169/300 4/4 [==============================] - 0s 10ms/step - loss: 0.7179 - val_loss: 0.6646 Epoch 170/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7176 - val_loss: 0.6643 Epoch 171/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7173 - val_loss: 0.6640 Epoch 172/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7170 - val_loss: 0.6638 Epoch 173/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7167 - val_loss: 0.6635 Epoch 174/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7164 - val_loss: 0.6631 Epoch 175/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7161 - val_loss: 0.6627 Epoch 176/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7158 - val_loss: 0.6623 Epoch 177/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7155 - val_loss: 0.6619 Epoch 178/300 4/4 [==============================] - 0s 8ms/step - loss: 0.7152 - val_loss: 0.6615 Epoch 179/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7149 - val_loss: 0.6612 Epoch 180/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7146 - val_loss: 0.6608 Epoch 181/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7143 - val_loss: 0.6605 Epoch 182/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7140 - val_loss: 0.6602 Epoch 183/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7137 - val_loss: 0.6600 Epoch 184/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7135 - val_loss: 0.6597 Epoch 185/300 4/4 [==============================] - 0s 10ms/step - loss: 0.7131 - val_loss: 0.6594 Epoch 186/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7128 - val_loss: 0.6591 Epoch 187/300 4/4 [==============================] - 0s 10ms/step - loss: 0.7126 - val_loss: 0.6588 Epoch 188/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7123 - val_loss: 0.6585 Epoch 189/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7120 - val_loss: 0.6583 Epoch 190/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7117 - val_loss: 0.6579 Epoch 191/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7114 - val_loss: 0.6576 Epoch 192/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7112 - val_loss: 0.6572 Epoch 193/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7109 - val_loss: 0.6569 Epoch 194/300 4/4 [==============================] - 0s 12ms/step - loss: 0.7106 - val_loss: 0.6565 Epoch 195/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7104 - val_loss: 0.6560 Epoch 196/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7101 - val_loss: 0.6556 Epoch 197/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7098 - val_loss: 0.6552 Epoch 198/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7095 - val_loss: 0.6548 Epoch 199/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7092 - val_loss: 0.6544 Epoch 200/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7089 - val_loss: 0.6540 Epoch 201/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7086 - val_loss: 0.6536 Epoch 202/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7083 - val_loss: 0.6532 Epoch 203/300 4/4 [==============================] - 0s 10ms/step - loss: 0.7079 - val_loss: 0.6527 Epoch 204/300 4/4 [==============================] - 0s 10ms/step - loss: 0.7076 - val_loss: 0.6523 Epoch 205/300 4/4 [==============================] - 0s 11ms/step - loss: 0.7073 - val_loss: 0.6520 Epoch 206/300 4/4 [==============================] - 0s 10ms/step - loss: 0.7070 - val_loss: 0.6516 Epoch 207/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7067 - val_loss: 0.6514 Epoch 208/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7064 - val_loss: 0.6511 Epoch 209/300 4/4 [==============================] - 0s 10ms/step - loss: 0.7061 - val_loss: 0.6509 Epoch 210/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7058 - val_loss: 0.6506 Epoch 211/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7055 - val_loss: 0.6505 Epoch 212/300 4/4 [==============================] - 0s 10ms/step - loss: 0.7053 - val_loss: 0.6502 Epoch 213/300 4/4 [==============================] - 0s 11ms/step - loss: 0.7050 - val_loss: 0.6500 Epoch 214/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7047 - val_loss: 0.6497 Epoch 215/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7045 - val_loss: 0.6494 Epoch 216/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7042 - val_loss: 0.6492 Epoch 217/300 4/4 [==============================] - 0s 11ms/step - loss: 0.7040 - val_loss: 0.6489 Epoch 218/300 4/4 [==============================] - 0s 10ms/step - loss: 0.7037 - val_loss: 0.6487 Epoch 219/300 4/4 [==============================] - 0s 11ms/step - loss: 0.7035 - val_loss: 0.6485 Epoch 220/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7032 - val_loss: 0.6483 Epoch 221/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7030 - val_loss: 0.6479 Epoch 222/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7027 - val_loss: 0.6477 Epoch 223/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7025 - val_loss: 0.6475 Epoch 224/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7022 - val_loss: 0.6472 Epoch 225/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7020 - val_loss: 0.6469 Epoch 226/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7017 - val_loss: 0.6465 Epoch 227/300 4/4 [==============================] - 0s 10ms/step - loss: 0.7015 - val_loss: 0.6461 Epoch 228/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7012 - val_loss: 0.6459 Epoch 229/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7010 - val_loss: 0.6456 Epoch 230/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7007 - val_loss: 0.6453 Epoch 231/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7004 - val_loss: 0.6450 Epoch 232/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7002 - val_loss: 0.6448 Epoch 233/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6999 - val_loss: 0.6446 Epoch 234/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6997 - val_loss: 0.6444 Epoch 235/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6995 - val_loss: 0.6441 Epoch 236/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6992 - val_loss: 0.6438 Epoch 237/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6990 - val_loss: 0.6435 Epoch 238/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6987 - val_loss: 0.6433 Epoch 239/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6985 - val_loss: 0.6430 Epoch 240/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6982 - val_loss: 0.6427 Epoch 241/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6979 - val_loss: 0.6424 Epoch 242/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6977 - val_loss: 0.6420 Epoch 243/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6975 - val_loss: 0.6418 Epoch 244/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6972 - val_loss: 0.6415 Epoch 245/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6970 - val_loss: 0.6412 Epoch 246/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6967 - val_loss: 0.6410 Epoch 247/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6964 - val_loss: 0.6406 Epoch 248/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6962 - val_loss: 0.6404 Epoch 249/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6959 - val_loss: 0.6401 Epoch 250/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6957 - val_loss: 0.6399 Epoch 251/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6954 - val_loss: 0.6395 Epoch 252/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6951 - val_loss: 0.6393 Epoch 253/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6949 - val_loss: 0.6390 Epoch 254/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6946 - val_loss: 0.6388 Epoch 255/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6944 - val_loss: 0.6385 Epoch 256/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6942 - val_loss: 0.6384 Epoch 257/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6940 - val_loss: 0.6382 Epoch 258/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6937 - val_loss: 0.6378 Epoch 259/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6934 - val_loss: 0.6374 Epoch 260/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6932 - val_loss: 0.6372 Epoch 261/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6930 - val_loss: 0.6368 Epoch 262/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6927 - val_loss: 0.6365 Epoch 263/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6925 - val_loss: 0.6362 Epoch 264/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6922 - val_loss: 0.6358 Epoch 265/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6920 - val_loss: 0.6356 Epoch 266/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6918 - val_loss: 0.6353 Epoch 267/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6915 - val_loss: 0.6351 Epoch 268/300 4/4 [==============================] - 0s 8ms/step - loss: 0.6913 - val_loss: 0.6347 Epoch 269/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6910 - val_loss: 0.6343 Epoch 270/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6908 - val_loss: 0.6341 Epoch 271/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6905 - val_loss: 0.6337 Epoch 272/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6902 - val_loss: 0.6334 Epoch 273/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6900 - val_loss: 0.6332 Epoch 274/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6898 - val_loss: 0.6328 Epoch 275/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6895 - val_loss: 0.6326 Epoch 276/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6892 - val_loss: 0.6324 Epoch 277/300 4/4 [==============================] - 0s 16ms/step - loss: 0.6890 - val_loss: 0.6321 Epoch 278/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6887 - val_loss: 0.6317 Epoch 279/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6885 - val_loss: 0.6314 Epoch 280/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6882 - val_loss: 0.6311 Epoch 281/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6880 - val_loss: 0.6306 Epoch 282/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6877 - val_loss: 0.6303 Epoch 283/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6874 - val_loss: 0.6301 Epoch 284/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6872 - val_loss: 0.6297 Epoch 285/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6869 - val_loss: 0.6295 Epoch 286/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6867 - val_loss: 0.6291 Epoch 287/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6864 - val_loss: 0.6289 Epoch 288/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6861 - val_loss: 0.6286 Epoch 289/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6859 - val_loss: 0.6282 Epoch 290/300 4/4 [==============================] - 0s 15ms/step - loss: 0.6856 - val_loss: 0.6278 Epoch 291/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6853 - val_loss: 0.6273 Epoch 292/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6850 - val_loss: 0.6271 Epoch 293/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6847 - val_loss: 0.6265 Epoch 294/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6844 - val_loss: 0.6261 Epoch 295/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6841 - val_loss: 0.6258 Epoch 296/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6838 - val_loss: 0.6253 Epoch 297/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6835 - val_loss: 0.6248 Epoch 298/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6832 - val_loss: 0.6245 Epoch 299/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6830 - val_loss: 0.6241 Epoch 300/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6827 - val_loss: 0.6235 Epoch 1/300 4/4 [==============================] - 0s 17ms/step - loss: 1.0752 - val_loss: 1.0588 Epoch 2/300 4/4 [==============================] - 0s 9ms/step - loss: 1.0625 - val_loss: 1.0504 Epoch 3/300 4/4 [==============================] - 0s 10ms/step - loss: 1.0541 - val_loss: 1.0436 Epoch 4/300 4/4 [==============================] - 0s 9ms/step - loss: 1.0472 - val_loss: 1.0379 Epoch 5/300 4/4 [==============================] - 0s 9ms/step - loss: 1.0413 - val_loss: 1.0328 Epoch 6/300 4/4 [==============================] - 0s 9ms/step - loss: 1.0362 - val_loss: 1.0283 Epoch 7/300 4/4 [==============================] - 0s 9ms/step - loss: 1.0315 - val_loss: 1.0240 Epoch 8/300 4/4 [==============================] - 0s 9ms/step - loss: 1.0272 - val_loss: 1.0203 Epoch 9/300 4/4 [==============================] - 0s 9ms/step - loss: 1.0231 - val_loss: 1.0165 Epoch 10/300 4/4 [==============================] - 0s 10ms/step - loss: 1.0191 - val_loss: 1.0132 Epoch 11/300 4/4 [==============================] - 0s 11ms/step - loss: 1.0154 - val_loss: 1.0097 Epoch 12/300 4/4 [==============================] - 0s 9ms/step - loss: 1.0117 - val_loss: 1.0065 Epoch 13/300 4/4 [==============================] - 0s 10ms/step - loss: 1.0082 - val_loss: 1.0033 Epoch 14/300 4/4 [==============================] - 0s 9ms/step - loss: 1.0049 - val_loss: 0.9999 Epoch 15/300 4/4 [==============================] - 0s 9ms/step - loss: 1.0015 - val_loss: 0.9972 Epoch 16/300 4/4 [==============================] - 0s 9ms/step - loss: 0.9983 - val_loss: 0.9941 Epoch 17/300 4/4 [==============================] - 0s 11ms/step - loss: 0.9951 - val_loss: 0.9914 Epoch 18/300 4/4 [==============================] - 0s 9ms/step - loss: 0.9921 - val_loss: 0.9883 Epoch 19/300 4/4 [==============================] - 0s 12ms/step - loss: 0.9891 - val_loss: 0.9855 Epoch 20/300 4/4 [==============================] - 0s 9ms/step - loss: 0.9861 - val_loss: 0.9823 Epoch 21/300 4/4 [==============================] - 0s 11ms/step - loss: 0.9831 - val_loss: 0.9797 Epoch 22/300 4/4 [==============================] - 0s 9ms/step - loss: 0.9803 - val_loss: 0.9767 Epoch 23/300 4/4 [==============================] - 0s 10ms/step - loss: 0.9774 - val_loss: 0.9738 Epoch 24/300 4/4 [==============================] - 0s 10ms/step - loss: 0.9745 - val_loss: 0.9708 Epoch 25/300 4/4 [==============================] - 0s 10ms/step - loss: 0.9715 - val_loss: 0.9680 Epoch 26/300 4/4 [==============================] - 0s 10ms/step - loss: 0.9686 - val_loss: 0.9651 Epoch 27/300 4/4 [==============================] - 0s 10ms/step - loss: 0.9656 - val_loss: 0.9622 Epoch 28/300 4/4 [==============================] - 0s 9ms/step - loss: 0.9627 - val_loss: 0.9592 Epoch 29/300 4/4 [==============================] - 0s 10ms/step - loss: 0.9597 - val_loss: 0.9567 Epoch 30/300 4/4 [==============================] - 0s 11ms/step - loss: 0.9567 - val_loss: 0.9535 Epoch 31/300 4/4 [==============================] - 0s 9ms/step - loss: 0.9537 - val_loss: 0.9504 Epoch 32/300 4/4 [==============================] - 0s 9ms/step - loss: 0.9507 - val_loss: 0.9474 Epoch 33/300 4/4 [==============================] - 0s 10ms/step - loss: 0.9477 - val_loss: 0.9442 Epoch 34/300 4/4 [==============================] - 0s 10ms/step - loss: 0.9447 - val_loss: 0.9414 Epoch 35/300 4/4 [==============================] - 0s 8ms/step - loss: 0.9416 - val_loss: 0.9384 Epoch 36/300 4/4 [==============================] - 0s 10ms/step - loss: 0.9386 - val_loss: 0.9353 Epoch 37/300 4/4 [==============================] - 0s 9ms/step - loss: 0.9354 - val_loss: 0.9327 Epoch 38/300 4/4 [==============================] - 0s 9ms/step - loss: 0.9324 - val_loss: 0.9294 Epoch 39/300 4/4 [==============================] - 0s 9ms/step - loss: 0.9292 - val_loss: 0.9264 Epoch 40/300 4/4 [==============================] - 0s 9ms/step - loss: 0.9262 - val_loss: 0.9232 Epoch 41/300 4/4 [==============================] - 0s 15ms/step - loss: 0.9230 - val_loss: 0.9203 Epoch 42/300 4/4 [==============================] - 0s 14ms/step - loss: 0.9199 - val_loss: 0.9169 Epoch 43/300 4/4 [==============================] - 0s 10ms/step - loss: 0.9167 - val_loss: 0.9138 Epoch 44/300 4/4 [==============================] - 0s 9ms/step - loss: 0.9135 - val_loss: 0.9104 Epoch 45/300 4/4 [==============================] - 0s 11ms/step - loss: 0.9103 - val_loss: 0.9075 Epoch 46/300 4/4 [==============================] - 0s 10ms/step - loss: 0.9071 - val_loss: 0.9044 Epoch 47/300 4/4 [==============================] - 0s 9ms/step - loss: 0.9039 - val_loss: 0.9011 Epoch 48/300 4/4 [==============================] - 0s 9ms/step - loss: 0.9006 - val_loss: 0.8979 Epoch 49/300 4/4 [==============================] - 0s 10ms/step - loss: 0.8973 - val_loss: 0.8944 Epoch 50/300 4/4 [==============================] - 0s 10ms/step - loss: 0.8939 - val_loss: 0.8909 Epoch 51/300 4/4 [==============================] - 0s 10ms/step - loss: 0.8906 - val_loss: 0.8877 Epoch 52/300 4/4 [==============================] - 0s 10ms/step - loss: 0.8873 - val_loss: 0.8844 Epoch 53/300 4/4 [==============================] - 0s 10ms/step - loss: 0.8840 - val_loss: 0.8811 Epoch 54/300 4/4 [==============================] - 0s 10ms/step - loss: 0.8806 - val_loss: 0.8779 Epoch 55/300 4/4 [==============================] - 0s 9ms/step - loss: 0.8772 - val_loss: 0.8745 Epoch 56/300 4/4 [==============================] - 0s 11ms/step - loss: 0.8738 - val_loss: 0.8713 Epoch 57/300 4/4 [==============================] - 0s 9ms/step - loss: 0.8705 - val_loss: 0.8679 Epoch 58/300 4/4 [==============================] - 0s 9ms/step - loss: 0.8670 - val_loss: 0.8643 Epoch 59/300 4/4 [==============================] - 0s 9ms/step - loss: 0.8635 - val_loss: 0.8610 Epoch 60/300 4/4 [==============================] - 0s 10ms/step - loss: 0.8602 - val_loss: 0.8574 Epoch 61/300 4/4 [==============================] - 0s 9ms/step - loss: 0.8566 - val_loss: 0.8543 Epoch 62/300 4/4 [==============================] - 0s 10ms/step - loss: 0.8532 - val_loss: 0.8512 Epoch 63/300 4/4 [==============================] - 0s 10ms/step - loss: 0.8496 - val_loss: 0.8475 Epoch 64/300 4/4 [==============================] - 0s 9ms/step - loss: 0.8461 - val_loss: 0.8441 Epoch 65/300 4/4 [==============================] - 0s 12ms/step - loss: 0.8425 - val_loss: 0.8408 Epoch 66/300 4/4 [==============================] - 0s 10ms/step - loss: 0.8390 - val_loss: 0.8374 Epoch 67/300 4/4 [==============================] - 0s 10ms/step - loss: 0.8355 - val_loss: 0.8340 Epoch 68/300 4/4 [==============================] - 0s 9ms/step - loss: 0.8319 - val_loss: 0.8309 Epoch 69/300 4/4 [==============================] - 0s 10ms/step - loss: 0.8284 - val_loss: 0.8271 Epoch 70/300 4/4 [==============================] - 0s 10ms/step - loss: 0.8249 - val_loss: 0.8240 Epoch 71/300 4/4 [==============================] - 0s 9ms/step - loss: 0.8215 - val_loss: 0.8208 Epoch 72/300 4/4 [==============================] - 0s 9ms/step - loss: 0.8181 - val_loss: 0.8172 Epoch 73/300 4/4 [==============================] - 0s 11ms/step - loss: 0.8146 - val_loss: 0.8134 Epoch 74/300 4/4 [==============================] - 0s 9ms/step - loss: 0.8110 - val_loss: 0.8102 Epoch 75/300 4/4 [==============================] - 0s 11ms/step - loss: 0.8076 - val_loss: 0.8067 Epoch 76/300 4/4 [==============================] - 0s 16ms/step - loss: 0.8041 - val_loss: 0.8031 Epoch 77/300 4/4 [==============================] - 0s 9ms/step - loss: 0.8007 - val_loss: 0.8001 Epoch 78/300 4/4 [==============================] - 0s 10ms/step - loss: 0.7974 - val_loss: 0.7967 Epoch 79/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7940 - val_loss: 0.7933 Epoch 80/300 4/4 [==============================] - 0s 10ms/step - loss: 0.7905 - val_loss: 0.7898 Epoch 81/300 4/4 [==============================] - 0s 10ms/step - loss: 0.7870 - val_loss: 0.7860 Epoch 82/300 4/4 [==============================] - 0s 10ms/step - loss: 0.7835 - val_loss: 0.7832 Epoch 83/300 4/4 [==============================] - 0s 12ms/step - loss: 0.7802 - val_loss: 0.7794 Epoch 84/300 4/4 [==============================] - 0s 10ms/step - loss: 0.7768 - val_loss: 0.7757 Epoch 85/300 4/4 [==============================] - 0s 10ms/step - loss: 0.7734 - val_loss: 0.7727 Epoch 86/300 4/4 [==============================] - 0s 10ms/step - loss: 0.7702 - val_loss: 0.7694 Epoch 87/300 4/4 [==============================] - 0s 10ms/step - loss: 0.7668 - val_loss: 0.7660 Epoch 88/300 4/4 [==============================] - 0s 12ms/step - loss: 0.7636 - val_loss: 0.7631 Epoch 89/300 4/4 [==============================] - 0s 10ms/step - loss: 0.7604 - val_loss: 0.7596 Epoch 90/300 4/4 [==============================] - 0s 10ms/step - loss: 0.7572 - val_loss: 0.7557 Epoch 91/300 4/4 [==============================] - 0s 10ms/step - loss: 0.7539 - val_loss: 0.7531 Epoch 92/300 4/4 [==============================] - 0s 10ms/step - loss: 0.7509 - val_loss: 0.7496 Epoch 93/300 4/4 [==============================] - 0s 10ms/step - loss: 0.7479 - val_loss: 0.7467 Epoch 94/300 4/4 [==============================] - 0s 10ms/step - loss: 0.7449 - val_loss: 0.7440 Epoch 95/300 4/4 [==============================] - 0s 10ms/step - loss: 0.7419 - val_loss: 0.7413 Epoch 96/300 4/4 [==============================] - 0s 13ms/step - loss: 0.7390 - val_loss: 0.7381 Epoch 97/300 4/4 [==============================] - 0s 10ms/step - loss: 0.7361 - val_loss: 0.7350 Epoch 98/300 4/4 [==============================] - 0s 10ms/step - loss: 0.7332 - val_loss: 0.7324 Epoch 99/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7305 - val_loss: 0.7300 Epoch 100/300 4/4 [==============================] - 0s 10ms/step - loss: 0.7279 - val_loss: 0.7273 Epoch 101/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7254 - val_loss: 0.7244 Epoch 102/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7228 - val_loss: 0.7221 Epoch 103/300 4/4 [==============================] - 0s 11ms/step - loss: 0.7203 - val_loss: 0.7199 Epoch 104/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7178 - val_loss: 0.7176 Epoch 105/300 4/4 [==============================] - 0s 10ms/step - loss: 0.7154 - val_loss: 0.7155 Epoch 106/300 4/4 [==============================] - 0s 10ms/step - loss: 0.7130 - val_loss: 0.7133 Epoch 107/300 4/4 [==============================] - 0s 10ms/step - loss: 0.7107 - val_loss: 0.7111 Epoch 108/300 4/4 [==============================] - 0s 10ms/step - loss: 0.7084 - val_loss: 0.7092 Epoch 109/300 4/4 [==============================] - 0s 10ms/step - loss: 0.7061 - val_loss: 0.7076 Epoch 110/300 4/4 [==============================] - 0s 10ms/step - loss: 0.7039 - val_loss: 0.7058 Epoch 111/300 4/4 [==============================] - 0s 10ms/step - loss: 0.7018 - val_loss: 0.7036 Epoch 112/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6998 - val_loss: 0.7028 Epoch 113/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6980 - val_loss: 0.7013 Epoch 114/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6962 - val_loss: 0.7002 Epoch 115/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6945 - val_loss: 0.6993 Epoch 116/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6929 - val_loss: 0.6983 Epoch 117/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6913 - val_loss: 0.6970 Epoch 118/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6899 - val_loss: 0.6962 Epoch 119/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6885 - val_loss: 0.6952 Epoch 120/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6872 - val_loss: 0.6944 Epoch 121/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6859 - val_loss: 0.6935 Epoch 122/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6845 - val_loss: 0.6925 Epoch 123/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6833 - val_loss: 0.6919 Epoch 124/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6822 - val_loss: 0.6915 Epoch 125/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6812 - val_loss: 0.6906 Epoch 126/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6801 - val_loss: 0.6900 Epoch 127/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6792 - val_loss: 0.6895 Epoch 128/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6784 - val_loss: 0.6894 Epoch 129/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6775 - val_loss: 0.6891 Epoch 130/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6767 - val_loss: 0.6886 Epoch 131/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6760 - val_loss: 0.6886 Epoch 132/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6754 - val_loss: 0.6880 Epoch 133/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6748 - val_loss: 0.6877 Epoch 134/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6741 - val_loss: 0.6879 Epoch 135/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6737 - val_loss: 0.6876 Epoch 136/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6730 - val_loss: 0.6874 Epoch 137/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6725 - val_loss: 0.6867 Epoch 138/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6720 - val_loss: 0.6869 Epoch 139/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6716 - val_loss: 0.6863 Epoch 140/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6711 - val_loss: 0.6858 Epoch 141/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6707 - val_loss: 0.6856 Epoch 142/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6702 - val_loss: 0.6846 Epoch 143/300 4/4 [==============================] - 0s 16ms/step - loss: 0.6698 - val_loss: 0.6849 Epoch 144/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6694 - val_loss: 0.6845 Epoch 145/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6689 - val_loss: 0.6839 Epoch 146/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6684 - val_loss: 0.6837 Epoch 147/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6683 - val_loss: 0.6831 Epoch 148/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6677 - val_loss: 0.6829 Epoch 149/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6672 - val_loss: 0.6823 Epoch 150/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6669 - val_loss: 0.6817 Epoch 151/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6665 - val_loss: 0.6811 Epoch 152/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6661 - val_loss: 0.6810 Epoch 153/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6657 - val_loss: 0.6808 Epoch 154/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6652 - val_loss: 0.6803 Epoch 155/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6648 - val_loss: 0.6803 Epoch 156/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6646 - val_loss: 0.6797 Epoch 157/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6641 - val_loss: 0.6792 Epoch 158/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6638 - val_loss: 0.6793 Epoch 159/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6633 - val_loss: 0.6788 Epoch 160/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6630 - val_loss: 0.6784 Epoch 161/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6627 - val_loss: 0.6779 Epoch 162/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6624 - val_loss: 0.6774 Epoch 163/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6620 - val_loss: 0.6770 Epoch 164/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6616 - val_loss: 0.6767 Epoch 165/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6614 - val_loss: 0.6764 Epoch 166/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6610 - val_loss: 0.6763 Epoch 167/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6606 - val_loss: 0.6758 Epoch 168/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6603 - val_loss: 0.6749 Epoch 169/300 4/4 [==============================] - 0s 16ms/step - loss: 0.6599 - val_loss: 0.6747 Epoch 170/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6597 - val_loss: 0.6747 Epoch 171/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6594 - val_loss: 0.6738 Epoch 172/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6590 - val_loss: 0.6738 Epoch 173/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6588 - val_loss: 0.6735 Epoch 174/300 4/4 [==============================] - 0s 16ms/step - loss: 0.6585 - val_loss: 0.6729 Epoch 175/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6581 - val_loss: 0.6731 Epoch 176/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6578 - val_loss: 0.6728 Epoch 177/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6575 - val_loss: 0.6723 Epoch 178/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6572 - val_loss: 0.6721 Epoch 179/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6569 - val_loss: 0.6720 Epoch 180/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6567 - val_loss: 0.6714 Epoch 181/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6564 - val_loss: 0.6711 Epoch 182/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6561 - val_loss: 0.6708 Epoch 183/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6557 - val_loss: 0.6706 Epoch 184/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6555 - val_loss: 0.6705 Epoch 185/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6553 - val_loss: 0.6701 Epoch 186/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6551 - val_loss: 0.6692 Epoch 187/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6547 - val_loss: 0.6691 Epoch 188/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6544 - val_loss: 0.6687 Epoch 189/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6542 - val_loss: 0.6686 Epoch 190/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6540 - val_loss: 0.6679 Epoch 191/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6537 - val_loss: 0.6674 Epoch 192/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6535 - val_loss: 0.6675 Epoch 193/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6532 - val_loss: 0.6672 Epoch 194/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6530 - val_loss: 0.6670 Epoch 195/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6527 - val_loss: 0.6665 Epoch 196/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6525 - val_loss: 0.6666 Epoch 197/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6522 - val_loss: 0.6662 Epoch 198/300 4/4 [==============================] - 0s 14ms/step - loss: 0.6520 - val_loss: 0.6661 Epoch 199/300 4/4 [==============================] - 0s 13ms/step - loss: 0.6518 - val_loss: 0.6655 Epoch 200/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6515 - val_loss: 0.6649 Epoch 201/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6512 - val_loss: 0.6653 Epoch 202/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6510 - val_loss: 0.6651 Epoch 203/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6507 - val_loss: 0.6645 Epoch 204/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6506 - val_loss: 0.6644 Epoch 205/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6503 - val_loss: 0.6638 Epoch 206/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6501 - val_loss: 0.6640 Epoch 207/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6499 - val_loss: 0.6635 Epoch 208/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6496 - val_loss: 0.6635 Epoch 209/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6495 - val_loss: 0.6631 Epoch 210/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6492 - val_loss: 0.6627 Epoch 211/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6491 - val_loss: 0.6628 Epoch 212/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6489 - val_loss: 0.6627 Epoch 213/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6486 - val_loss: 0.6623 Epoch 214/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6486 - val_loss: 0.6619 Epoch 215/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6483 - val_loss: 0.6617 Epoch 216/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6481 - val_loss: 0.6616 Epoch 217/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6480 - val_loss: 0.6616 Epoch 218/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6477 - val_loss: 0.6613 Epoch 219/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6475 - val_loss: 0.6614 Epoch 220/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6473 - val_loss: 0.6610 Epoch 221/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6471 - val_loss: 0.6607 Epoch 222/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6469 - val_loss: 0.6604 Epoch 223/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6468 - val_loss: 0.6605 Epoch 224/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6467 - val_loss: 0.6602 Epoch 225/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6465 - val_loss: 0.6597 Epoch 226/300 4/4 [==============================] - 0s 16ms/step - loss: 0.6462 - val_loss: 0.6593 Epoch 227/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6460 - val_loss: 0.6593 Epoch 228/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6459 - val_loss: 0.6594 Epoch 229/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6458 - val_loss: 0.6593 Epoch 230/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6455 - val_loss: 0.6591 Epoch 231/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6454 - val_loss: 0.6590 Epoch 232/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6452 - val_loss: 0.6590 Epoch 233/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6451 - val_loss: 0.6585 Epoch 234/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6450 - val_loss: 0.6588 Epoch 235/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6448 - val_loss: 0.6583 Epoch 236/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6447 - val_loss: 0.6582 Epoch 237/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6446 - val_loss: 0.6582 Epoch 238/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6444 - val_loss: 0.6579 Epoch 239/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6443 - val_loss: 0.6570 Epoch 240/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6440 - val_loss: 0.6573 Epoch 241/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6440 - val_loss: 0.6571 Epoch 242/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6438 - val_loss: 0.6569 Epoch 243/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6437 - val_loss: 0.6567 Epoch 244/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6435 - val_loss: 0.6565 Epoch 245/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6435 - val_loss: 0.6562 Epoch 246/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6433 - val_loss: 0.6564 Epoch 247/300 4/4 [==============================] - 0s 13ms/step - loss: 0.6431 - val_loss: 0.6562 Epoch 248/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6431 - val_loss: 0.6560 Epoch 249/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6429 - val_loss: 0.6560 Epoch 250/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6428 - val_loss: 0.6556 Epoch 251/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6426 - val_loss: 0.6556 Epoch 252/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6425 - val_loss: 0.6555 Epoch 253/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6424 - val_loss: 0.6555 Epoch 254/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6423 - val_loss: 0.6552 Epoch 255/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6422 - val_loss: 0.6549 Epoch 256/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6421 - val_loss: 0.6548 Epoch 257/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6419 - val_loss: 0.6548 Epoch 258/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6417 - val_loss: 0.6546 Epoch 259/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6417 - val_loss: 0.6545 Epoch 260/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6416 - val_loss: 0.6547 Epoch 261/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6416 - val_loss: 0.6545 Epoch 262/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6414 - val_loss: 0.6541 Epoch 263/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6413 - val_loss: 0.6541 Epoch 264/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6412 - val_loss: 0.6533 Epoch 265/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6411 - val_loss: 0.6535 Epoch 266/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6410 - val_loss: 0.6531 Epoch 267/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6408 - val_loss: 0.6535 Epoch 268/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6408 - val_loss: 0.6532 Epoch 269/300 4/4 [==============================] - 0s 19ms/step - loss: 0.6406 - val_loss: 0.6529 Epoch 270/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6406 - val_loss: 0.6528 Epoch 271/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6404 - val_loss: 0.6524 Epoch 272/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6404 - val_loss: 0.6524 Epoch 273/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6403 - val_loss: 0.6524 Epoch 274/300 4/4 [==============================] - 0s 16ms/step - loss: 0.6402 - val_loss: 0.6522 Epoch 275/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6401 - val_loss: 0.6521 Epoch 276/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6400 - val_loss: 0.6522 Epoch 277/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6399 - val_loss: 0.6517 Epoch 278/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6398 - val_loss: 0.6519 Epoch 279/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6397 - val_loss: 0.6514 Epoch 280/300 4/4 [==============================] - 0s 14ms/step - loss: 0.6396 - val_loss: 0.6517 Epoch 281/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6397 - val_loss: 0.6511 Epoch 282/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6394 - val_loss: 0.6516 Epoch 283/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6394 - val_loss: 0.6511 Epoch 284/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6393 - val_loss: 0.6511 Epoch 285/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6392 - val_loss: 0.6508 Epoch 286/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6391 - val_loss: 0.6506 Epoch 287/300 4/4 [==============================] - 0s 17ms/step - loss: 0.6390 - val_loss: 0.6506 Epoch 288/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6390 - val_loss: 0.6502 Epoch 289/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6388 - val_loss: 0.6505 Epoch 290/300 4/4 [==============================] - 0s 14ms/step - loss: 0.6388 - val_loss: 0.6504 Epoch 291/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6387 - val_loss: 0.6501 Epoch 292/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6386 - val_loss: 0.6503 Epoch 293/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6386 - val_loss: 0.6502 Epoch 294/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6385 - val_loss: 0.6499 Epoch 295/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6384 - val_loss: 0.6500 Epoch 296/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6384 - val_loss: 0.6496 Epoch 297/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6383 - val_loss: 0.6496 Epoch 298/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6382 - val_loss: 0.6494 Epoch 299/300 4/4 [==============================] - 0s 16ms/step - loss: 0.6381 - val_loss: 0.6493 Epoch 300/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6380 - val_loss: 0.6493 Epoch 1/300 4/4 [==============================] - 0s 19ms/step - loss: 1.0077 - val_loss: 1.0035 Epoch 2/300 4/4 [==============================] - 0s 10ms/step - loss: 0.9963 - val_loss: 0.9958 Epoch 3/300 4/4 [==============================] - 0s 10ms/step - loss: 0.9880 - val_loss: 0.9889 Epoch 4/300 4/4 [==============================] - 0s 10ms/step - loss: 0.9808 - val_loss: 0.9826 Epoch 5/300 4/4 [==============================] - 0s 10ms/step - loss: 0.9743 - val_loss: 0.9765 Epoch 6/300 4/4 [==============================] - 0s 11ms/step - loss: 0.9682 - val_loss: 0.9707 Epoch 7/300 4/4 [==============================] - 0s 11ms/step - loss: 0.9626 - val_loss: 0.9653 Epoch 8/300 4/4 [==============================] - 0s 11ms/step - loss: 0.9573 - val_loss: 0.9600 Epoch 9/300 4/4 [==============================] - 0s 11ms/step - loss: 0.9524 - val_loss: 0.9551 Epoch 10/300 4/4 [==============================] - 0s 12ms/step - loss: 0.9476 - val_loss: 0.9504 Epoch 11/300 4/4 [==============================] - 0s 10ms/step - loss: 0.9429 - val_loss: 0.9458 Epoch 12/300 4/4 [==============================] - 0s 10ms/step - loss: 0.9383 - val_loss: 0.9414 Epoch 13/300 4/4 [==============================] - 0s 10ms/step - loss: 0.9339 - val_loss: 0.9374 Epoch 14/300 4/4 [==============================] - 0s 11ms/step - loss: 0.9297 - val_loss: 0.9336 Epoch 15/300 4/4 [==============================] - 0s 10ms/step - loss: 0.9257 - val_loss: 0.9298 Epoch 16/300 4/4 [==============================] - 0s 10ms/step - loss: 0.9217 - val_loss: 0.9261 Epoch 17/300 4/4 [==============================] - 0s 11ms/step - loss: 0.9177 - val_loss: 0.9224 Epoch 18/300 4/4 [==============================] - 0s 10ms/step - loss: 0.9137 - val_loss: 0.9186 Epoch 19/300 4/4 [==============================] - 0s 10ms/step - loss: 0.9098 - val_loss: 0.9147 Epoch 20/300 4/4 [==============================] - 0s 10ms/step - loss: 0.9058 - val_loss: 0.9110 Epoch 21/300 4/4 [==============================] - 0s 10ms/step - loss: 0.9019 - val_loss: 0.9072 Epoch 22/300 4/4 [==============================] - 0s 10ms/step - loss: 0.8980 - val_loss: 0.9035 Epoch 23/300 4/4 [==============================] - 0s 11ms/step - loss: 0.8941 - val_loss: 0.8997 Epoch 24/300 4/4 [==============================] - 0s 11ms/step - loss: 0.8902 - val_loss: 0.8961 Epoch 25/300 4/4 [==============================] - 0s 10ms/step - loss: 0.8865 - val_loss: 0.8926 Epoch 26/300 4/4 [==============================] - 0s 10ms/step - loss: 0.8826 - val_loss: 0.8890 Epoch 27/300 4/4 [==============================] - 0s 9ms/step - loss: 0.8789 - val_loss: 0.8854 Epoch 28/300 4/4 [==============================] - 0s 10ms/step - loss: 0.8752 - val_loss: 0.8819 Epoch 29/300 4/4 [==============================] - 0s 11ms/step - loss: 0.8715 - val_loss: 0.8787 Epoch 30/300 4/4 [==============================] - 0s 10ms/step - loss: 0.8679 - val_loss: 0.8753 Epoch 31/300 4/4 [==============================] - 0s 10ms/step - loss: 0.8643 - val_loss: 0.8716 Epoch 32/300 4/4 [==============================] - 0s 12ms/step - loss: 0.8608 - val_loss: 0.8683 Epoch 33/300 4/4 [==============================] - 0s 10ms/step - loss: 0.8573 - val_loss: 0.8651 Epoch 34/300 4/4 [==============================] - 0s 10ms/step - loss: 0.8540 - val_loss: 0.8619 Epoch 35/300 4/4 [==============================] - 0s 11ms/step - loss: 0.8506 - val_loss: 0.8587 Epoch 36/300 4/4 [==============================] - 0s 10ms/step - loss: 0.8472 - val_loss: 0.8554 Epoch 37/300 4/4 [==============================] - 0s 9ms/step - loss: 0.8440 - val_loss: 0.8524 Epoch 38/300 4/4 [==============================] - 0s 11ms/step - loss: 0.8408 - val_loss: 0.8491 Epoch 39/300 4/4 [==============================] - 0s 10ms/step - loss: 0.8376 - val_loss: 0.8461 Epoch 40/300 4/4 [==============================] - 0s 9ms/step - loss: 0.8345 - val_loss: 0.8430 Epoch 41/300 4/4 [==============================] - 0s 11ms/step - loss: 0.8314 - val_loss: 0.8398 Epoch 42/300 4/4 [==============================] - 0s 11ms/step - loss: 0.8284 - val_loss: 0.8371 Epoch 43/300 4/4 [==============================] - 0s 11ms/step - loss: 0.8254 - val_loss: 0.8340 Epoch 44/300 4/4 [==============================] - 0s 12ms/step - loss: 0.8224 - val_loss: 0.8312 Epoch 45/300 4/4 [==============================] - 0s 9ms/step - loss: 0.8193 - val_loss: 0.8281 Epoch 46/300 4/4 [==============================] - 0s 9ms/step - loss: 0.8165 - val_loss: 0.8253 Epoch 47/300 4/4 [==============================] - 0s 10ms/step - loss: 0.8137 - val_loss: 0.8224 Epoch 48/300 4/4 [==============================] - 0s 10ms/step - loss: 0.8109 - val_loss: 0.8194 Epoch 49/300 4/4 [==============================] - 0s 9ms/step - loss: 0.8081 - val_loss: 0.8166 Epoch 50/300 4/4 [==============================] - 0s 10ms/step - loss: 0.8054 - val_loss: 0.8138 Epoch 51/300 4/4 [==============================] - 0s 10ms/step - loss: 0.8027 - val_loss: 0.8112 Epoch 52/300 4/4 [==============================] - 0s 10ms/step - loss: 0.8000 - val_loss: 0.8085 Epoch 53/300 4/4 [==============================] - 0s 11ms/step - loss: 0.7974 - val_loss: 0.8060 Epoch 54/300 4/4 [==============================] - 0s 11ms/step - loss: 0.7947 - val_loss: 0.8032 Epoch 55/300 4/4 [==============================] - 0s 13ms/step - loss: 0.7921 - val_loss: 0.8004 Epoch 56/300 4/4 [==============================] - 0s 10ms/step - loss: 0.7896 - val_loss: 0.7976 Epoch 57/300 4/4 [==============================] - 0s 11ms/step - loss: 0.7870 - val_loss: 0.7949 Epoch 58/300 4/4 [==============================] - 0s 10ms/step - loss: 0.7844 - val_loss: 0.7924 Epoch 59/300 4/4 [==============================] - 0s 10ms/step - loss: 0.7819 - val_loss: 0.7900 Epoch 60/300 4/4 [==============================] - 0s 11ms/step - loss: 0.7795 - val_loss: 0.7874 Epoch 61/300 4/4 [==============================] - 0s 11ms/step - loss: 0.7771 - val_loss: 0.7849 Epoch 62/300 4/4 [==============================] - 0s 10ms/step - loss: 0.7747 - val_loss: 0.7826 Epoch 63/300 4/4 [==============================] - 0s 11ms/step - loss: 0.7723 - val_loss: 0.7800 Epoch 64/300 4/4 [==============================] - 0s 10ms/step - loss: 0.7700 - val_loss: 0.7777 Epoch 65/300 4/4 [==============================] - 0s 11ms/step - loss: 0.7676 - val_loss: 0.7752 Epoch 66/300 4/4 [==============================] - 0s 10ms/step - loss: 0.7654 - val_loss: 0.7728 Epoch 67/300 4/4 [==============================] - 0s 11ms/step - loss: 0.7632 - val_loss: 0.7703 Epoch 68/300 4/4 [==============================] - 0s 11ms/step - loss: 0.7610 - val_loss: 0.7680 Epoch 69/300 4/4 [==============================] - 0s 12ms/step - loss: 0.7589 - val_loss: 0.7657 Epoch 70/300 4/4 [==============================] - 0s 12ms/step - loss: 0.7567 - val_loss: 0.7632 Epoch 71/300 4/4 [==============================] - 0s 12ms/step - loss: 0.7546 - val_loss: 0.7610 Epoch 72/300 4/4 [==============================] - 0s 11ms/step - loss: 0.7524 - val_loss: 0.7587 Epoch 73/300 4/4 [==============================] - 0s 12ms/step - loss: 0.7504 - val_loss: 0.7563 Epoch 74/300 4/4 [==============================] - 0s 11ms/step - loss: 0.7482 - val_loss: 0.7542 Epoch 75/300 4/4 [==============================] - 0s 13ms/step - loss: 0.7461 - val_loss: 0.7520 Epoch 76/300 4/4 [==============================] - 0s 15ms/step - loss: 0.7441 - val_loss: 0.7496 Epoch 77/300 4/4 [==============================] - 0s 11ms/step - loss: 0.7419 - val_loss: 0.7473 Epoch 78/300 4/4 [==============================] - 0s 11ms/step - loss: 0.7399 - val_loss: 0.7451 Epoch 79/300 4/4 [==============================] - 0s 11ms/step - loss: 0.7379 - val_loss: 0.7428 Epoch 80/300 4/4 [==============================] - 0s 11ms/step - loss: 0.7358 - val_loss: 0.7407 Epoch 81/300 4/4 [==============================] - 0s 10ms/step - loss: 0.7339 - val_loss: 0.7385 Epoch 82/300 4/4 [==============================] - 0s 11ms/step - loss: 0.7318 - val_loss: 0.7365 Epoch 83/300 4/4 [==============================] - 0s 11ms/step - loss: 0.7298 - val_loss: 0.7345 Epoch 84/300 4/4 [==============================] - 0s 11ms/step - loss: 0.7279 - val_loss: 0.7324 Epoch 85/300 4/4 [==============================] - 0s 10ms/step - loss: 0.7260 - val_loss: 0.7306 Epoch 86/300 4/4 [==============================] - 0s 10ms/step - loss: 0.7240 - val_loss: 0.7286 Epoch 87/300 4/4 [==============================] - 0s 10ms/step - loss: 0.7222 - val_loss: 0.7266 Epoch 88/300 4/4 [==============================] - 0s 10ms/step - loss: 0.7203 - val_loss: 0.7247 Epoch 89/300 4/4 [==============================] - 0s 11ms/step - loss: 0.7183 - val_loss: 0.7226 Epoch 90/300 4/4 [==============================] - 0s 11ms/step - loss: 0.7164 - val_loss: 0.7207 Epoch 91/300 4/4 [==============================] - 0s 11ms/step - loss: 0.7146 - val_loss: 0.7189 Epoch 92/300 4/4 [==============================] - 0s 10ms/step - loss: 0.7127 - val_loss: 0.7169 Epoch 93/300 4/4 [==============================] - 0s 12ms/step - loss: 0.7108 - val_loss: 0.7151 Epoch 94/300 4/4 [==============================] - 0s 10ms/step - loss: 0.7089 - val_loss: 0.7132 Epoch 95/300 4/4 [==============================] - 0s 11ms/step - loss: 0.7071 - val_loss: 0.7111 Epoch 96/300 4/4 [==============================] - 0s 10ms/step - loss: 0.7053 - val_loss: 0.7092 Epoch 97/300 4/4 [==============================] - 0s 9ms/step - loss: 0.7036 - val_loss: 0.7074 Epoch 98/300 4/4 [==============================] - 0s 13ms/step - loss: 0.7018 - val_loss: 0.7056 Epoch 99/300 4/4 [==============================] - 0s 11ms/step - loss: 0.7001 - val_loss: 0.7038 Epoch 100/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6984 - val_loss: 0.7020 Epoch 101/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6968 - val_loss: 0.7002 Epoch 102/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6951 - val_loss: 0.6984 Epoch 103/300 4/4 [==============================] - 0s 17ms/step - loss: 0.6935 - val_loss: 0.6966 Epoch 104/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6919 - val_loss: 0.6949 Epoch 105/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6903 - val_loss: 0.6931 Epoch 106/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6887 - val_loss: 0.6912 Epoch 107/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6871 - val_loss: 0.6897 Epoch 108/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6856 - val_loss: 0.6877 Epoch 109/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6841 - val_loss: 0.6860 Epoch 110/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6825 - val_loss: 0.6847 Epoch 111/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6811 - val_loss: 0.6831 Epoch 112/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6796 - val_loss: 0.6812 Epoch 113/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6782 - val_loss: 0.6798 Epoch 114/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6768 - val_loss: 0.6784 Epoch 115/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6754 - val_loss: 0.6768 Epoch 116/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6740 - val_loss: 0.6753 Epoch 117/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6728 - val_loss: 0.6737 Epoch 118/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6713 - val_loss: 0.6722 Epoch 119/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6700 - val_loss: 0.6708 Epoch 120/300 4/4 [==============================] - 0s 13ms/step - loss: 0.6688 - val_loss: 0.6695 Epoch 121/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6675 - val_loss: 0.6684 Epoch 122/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6666 - val_loss: 0.6664 Epoch 123/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6652 - val_loss: 0.6650 Epoch 124/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6641 - val_loss: 0.6640 Epoch 125/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6631 - val_loss: 0.6623 Epoch 126/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6619 - val_loss: 0.6606 Epoch 127/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6609 - val_loss: 0.6599 Epoch 128/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6598 - val_loss: 0.6582 Epoch 129/300 4/4 [==============================] - 0s 18ms/step - loss: 0.6589 - val_loss: 0.6571 Epoch 130/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6577 - val_loss: 0.6556 Epoch 131/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6566 - val_loss: 0.6545 Epoch 132/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6559 - val_loss: 0.6532 Epoch 133/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6547 - val_loss: 0.6523 Epoch 134/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6541 - val_loss: 0.6516 Epoch 135/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6535 - val_loss: 0.6503 Epoch 136/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6525 - val_loss: 0.6493 Epoch 137/300 4/4 [==============================] - 0s 13ms/step - loss: 0.6518 - val_loss: 0.6481 Epoch 138/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6508 - val_loss: 0.6470 Epoch 139/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6501 - val_loss: 0.6470 Epoch 140/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6497 - val_loss: 0.6457 Epoch 141/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6488 - val_loss: 0.6447 Epoch 142/300 4/4 [==============================] - 0s 13ms/step - loss: 0.6479 - val_loss: 0.6437 Epoch 143/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6473 - val_loss: 0.6429 Epoch 144/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6466 - val_loss: 0.6418 Epoch 145/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6459 - val_loss: 0.6410 Epoch 146/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6453 - val_loss: 0.6401 Epoch 147/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6446 - val_loss: 0.6391 Epoch 148/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6439 - val_loss: 0.6385 Epoch 149/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6435 - val_loss: 0.6378 Epoch 150/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6428 - val_loss: 0.6368 Epoch 151/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6420 - val_loss: 0.6365 Epoch 152/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6416 - val_loss: 0.6355 Epoch 153/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6411 - val_loss: 0.6348 Epoch 154/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6404 - val_loss: 0.6342 Epoch 155/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6399 - val_loss: 0.6335 Epoch 156/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6394 - val_loss: 0.6329 Epoch 157/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6388 - val_loss: 0.6323 Epoch 158/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6383 - val_loss: 0.6315 Epoch 159/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6377 - val_loss: 0.6308 Epoch 160/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6370 - val_loss: 0.6306 Epoch 161/300 4/4 [==============================] - 0s 17ms/step - loss: 0.6366 - val_loss: 0.6296 Epoch 162/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6359 - val_loss: 0.6291 Epoch 163/300 4/4 [==============================] - 0s 13ms/step - loss: 0.6354 - val_loss: 0.6281 Epoch 164/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6347 - val_loss: 0.6275 Epoch 165/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6343 - val_loss: 0.6276 Epoch 166/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6338 - val_loss: 0.6268 Epoch 167/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6331 - val_loss: 0.6259 Epoch 168/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6325 - val_loss: 0.6253 Epoch 169/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6319 - val_loss: 0.6246 Epoch 170/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6315 - val_loss: 0.6239 Epoch 171/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6308 - val_loss: 0.6236 Epoch 172/300 4/4 [==============================] - 0s 14ms/step - loss: 0.6303 - val_loss: 0.6227 Epoch 173/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6297 - val_loss: 0.6221 Epoch 174/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6291 - val_loss: 0.6217 Epoch 175/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6286 - val_loss: 0.6209 Epoch 176/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6281 - val_loss: 0.6198 Epoch 177/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6275 - val_loss: 0.6194 Epoch 178/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6269 - val_loss: 0.6187 Epoch 179/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6263 - val_loss: 0.6181 Epoch 180/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6258 - val_loss: 0.6178 Epoch 181/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6254 - val_loss: 0.6169 Epoch 182/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6248 - val_loss: 0.6166 Epoch 183/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6241 - val_loss: 0.6160 Epoch 184/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6237 - val_loss: 0.6153 Epoch 185/300 4/4 [==============================] - 0s 13ms/step - loss: 0.6230 - val_loss: 0.6152 Epoch 186/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6227 - val_loss: 0.6143 Epoch 187/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6221 - val_loss: 0.6137 Epoch 188/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6215 - val_loss: 0.6130 Epoch 189/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6210 - val_loss: 0.6125 Epoch 190/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6204 - val_loss: 0.6122 Epoch 191/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6199 - val_loss: 0.6117 Epoch 192/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6194 - val_loss: 0.6113 Epoch 193/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6189 - val_loss: 0.6105 Epoch 194/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6185 - val_loss: 0.6102 Epoch 195/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6178 - val_loss: 0.6096 Epoch 196/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6173 - val_loss: 0.6089 Epoch 197/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6168 - val_loss: 0.6088 Epoch 198/300 4/4 [==============================] - 0s 16ms/step - loss: 0.6163 - val_loss: 0.6079 Epoch 199/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6161 - val_loss: 0.6073 Epoch 200/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6155 - val_loss: 0.6068 Epoch 201/300 4/4 [==============================] - 0s 16ms/step - loss: 0.6151 - val_loss: 0.6067 Epoch 202/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6148 - val_loss: 0.6063 Epoch 203/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6145 - val_loss: 0.6060 Epoch 204/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6141 - val_loss: 0.6057 Epoch 205/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6139 - val_loss: 0.6052 Epoch 206/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6135 - val_loss: 0.6049 Epoch 207/300 4/4 [==============================] - 0s 13ms/step - loss: 0.6135 - val_loss: 0.6046 Epoch 208/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6131 - val_loss: 0.6040 Epoch 209/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6128 - val_loss: 0.6037 Epoch 210/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6126 - val_loss: 0.6036 Epoch 211/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6125 - val_loss: 0.6032 Epoch 212/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6124 - val_loss: 0.6032 Epoch 213/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6120 - val_loss: 0.6030 Epoch 214/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6122 - val_loss: 0.6027 Epoch 215/300 4/4 [==============================] - 0s 9ms/step - loss: 0.6119 - val_loss: 0.6026 Epoch 216/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6115 - val_loss: 0.6023 Epoch 217/300 4/4 [==============================] - 0s 13ms/step - loss: 0.6117 - val_loss: 0.6026 Epoch 218/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6116 - val_loss: 0.6020 Epoch 219/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6112 - val_loss: 0.6024 Epoch 220/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6113 - val_loss: 0.6019 Epoch 221/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6111 - val_loss: 0.6022 Epoch 222/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6110 - val_loss: 0.6019 Epoch 223/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6110 - val_loss: 0.6016 Epoch 224/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6106 - val_loss: 0.6019 Epoch 225/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6109 - val_loss: 0.6011 Epoch 226/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6107 - val_loss: 0.6013 Epoch 227/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6107 - val_loss: 0.6008 Epoch 228/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6105 - val_loss: 0.6010 Epoch 229/300 4/4 [==============================] - 0s 14ms/step - loss: 0.6106 - val_loss: 0.6008 Epoch 230/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6105 - val_loss: 0.6009 Epoch 231/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6103 - val_loss: 0.6005 Epoch 232/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6102 - val_loss: 0.6005 Epoch 233/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6102 - val_loss: 0.6006 Epoch 234/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6102 - val_loss: 0.6007 Epoch 235/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6102 - val_loss: 0.6003 Epoch 236/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6100 - val_loss: 0.6002 Epoch 237/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6100 - val_loss: 0.6000 Epoch 238/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6100 - val_loss: 0.6000 Epoch 239/300 4/4 [==============================] - 0s 13ms/step - loss: 0.6099 - val_loss: 0.5999 Epoch 240/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6099 - val_loss: 0.5999 Epoch 241/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6098 - val_loss: 0.5999 Epoch 242/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6098 - val_loss: 0.5996 Epoch 243/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6098 - val_loss: 0.5998 Epoch 244/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6097 - val_loss: 0.5994 Epoch 245/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6096 - val_loss: 0.5997 Epoch 246/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6096 - val_loss: 0.5995 Epoch 247/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6095 - val_loss: 0.5992 Epoch 248/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6095 - val_loss: 0.5995 Epoch 249/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6094 - val_loss: 0.5997 Epoch 250/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6093 - val_loss: 0.5992 Epoch 251/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6093 - val_loss: 0.5994 Epoch 252/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6094 - val_loss: 0.5988 Epoch 253/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6092 - val_loss: 0.5988 Epoch 254/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6092 - val_loss: 0.5989 Epoch 255/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6092 - val_loss: 0.5989 Epoch 256/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6092 - val_loss: 0.5987 Epoch 257/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6091 - val_loss: 0.5987 Epoch 258/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6090 - val_loss: 0.5989 Epoch 259/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6091 - val_loss: 0.5985 Epoch 260/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6090 - val_loss: 0.5981 Epoch 261/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6089 - val_loss: 0.5983 Epoch 262/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6090 - val_loss: 0.5983 Epoch 263/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6089 - val_loss: 0.5981 Epoch 264/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6089 - val_loss: 0.5980 Epoch 265/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6088 - val_loss: 0.5979 Epoch 266/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6088 - val_loss: 0.5979 Epoch 267/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6089 - val_loss: 0.5980 Epoch 268/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6087 - val_loss: 0.5982 Epoch 269/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6087 - val_loss: 0.5981 Epoch 270/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6088 - val_loss: 0.5981 Epoch 271/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6086 - val_loss: 0.5983 Epoch 272/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6087 - val_loss: 0.5979 Epoch 273/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6086 - val_loss: 0.5984 Epoch 274/300 4/4 [==============================] - 0s 14ms/step - loss: 0.6086 - val_loss: 0.5982 Epoch 275/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6086 - val_loss: 0.5979 Epoch 276/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6085 - val_loss: 0.5979 Epoch 277/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6085 - val_loss: 0.5974 Epoch 278/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6085 - val_loss: 0.5981 Epoch 279/300 4/4 [==============================] - 0s 13ms/step - loss: 0.6085 - val_loss: 0.5978 Epoch 280/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6085 - val_loss: 0.5976 Epoch 281/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6084 - val_loss: 0.5979 Epoch 282/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6085 - val_loss: 0.5973 Epoch 283/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6083 - val_loss: 0.5975 Epoch 284/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6083 - val_loss: 0.5978 Epoch 285/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6083 - val_loss: 0.5975 Epoch 286/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6083 - val_loss: 0.5976 Epoch 287/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6083 - val_loss: 0.5976 Epoch 288/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6083 - val_loss: 0.5974 Epoch 289/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6082 - val_loss: 0.5978 Epoch 290/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6083 - val_loss: 0.5972 Epoch 291/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6082 - val_loss: 0.5972 Epoch 292/300 4/4 [==============================] - 0s 14ms/step - loss: 0.6082 - val_loss: 0.5970 Epoch 293/300 4/4 [==============================] - 0s 13ms/step - loss: 0.6082 - val_loss: 0.5966 Epoch 294/300 4/4 [==============================] - 0s 14ms/step - loss: 0.6080 - val_loss: 0.5970 Epoch 295/300 4/4 [==============================] - 0s 13ms/step - loss: 0.6082 - val_loss: 0.5964 Epoch 296/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6081 - val_loss: 0.5970 Epoch 297/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6081 - val_loss: 0.5970 Epoch 298/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6080 - val_loss: 0.5968 Epoch 299/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6080 - val_loss: 0.5972 Epoch 300/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6081 - val_loss: 0.5968 Epoch 1/300 4/4 [==============================] - 0s 21ms/step - loss: 1.0347 - val_loss: 1.0561 Epoch 2/300 4/4 [==============================] - 0s 11ms/step - loss: 1.0303 - val_loss: 1.0524 Epoch 3/300 4/4 [==============================] - 0s 11ms/step - loss: 1.0269 - val_loss: 1.0499 Epoch 4/300 4/4 [==============================] - 0s 11ms/step - loss: 1.0243 - val_loss: 1.0475 Epoch 5/300 4/4 [==============================] - 0s 12ms/step - loss: 1.0217 - val_loss: 1.0455 Epoch 6/300 4/4 [==============================] - 0s 11ms/step - loss: 1.0193 - val_loss: 1.0435 Epoch 7/300 4/4 [==============================] - 0s 12ms/step - loss: 1.0171 - val_loss: 1.0415 Epoch 8/300 4/4 [==============================] - 0s 13ms/step - loss: 1.0149 - val_loss: 1.0395 Epoch 9/300 4/4 [==============================] - 0s 18ms/step - loss: 1.0127 - val_loss: 1.0375 Epoch 10/300 4/4 [==============================] - 0s 13ms/step - loss: 1.0106 - val_loss: 1.0358 Epoch 11/300 4/4 [==============================] - 0s 12ms/step - loss: 1.0085 - val_loss: 1.0341 Epoch 12/300 4/4 [==============================] - 0s 10ms/step - loss: 1.0067 - val_loss: 1.0322 Epoch 13/300 4/4 [==============================] - 0s 13ms/step - loss: 1.0047 - val_loss: 1.0308 Epoch 14/300 4/4 [==============================] - 0s 15ms/step - loss: 1.0027 - val_loss: 1.0288 Epoch 15/300 4/4 [==============================] - 0s 10ms/step - loss: 1.0008 - val_loss: 1.0271 Epoch 16/300 4/4 [==============================] - 0s 10ms/step - loss: 0.9988 - val_loss: 1.0256 Epoch 17/300 4/4 [==============================] - 0s 11ms/step - loss: 0.9970 - val_loss: 1.0238 Epoch 18/300 4/4 [==============================] - 0s 12ms/step - loss: 0.9951 - val_loss: 1.0224 Epoch 19/300 4/4 [==============================] - 0s 10ms/step - loss: 0.9933 - val_loss: 1.0205 Epoch 20/300 4/4 [==============================] - 0s 11ms/step - loss: 0.9915 - val_loss: 1.0190 Epoch 21/300 4/4 [==============================] - 0s 11ms/step - loss: 0.9896 - val_loss: 1.0174 Epoch 22/300 4/4 [==============================] - 0s 10ms/step - loss: 0.9880 - val_loss: 1.0156 Epoch 23/300 4/4 [==============================] - 0s 11ms/step - loss: 0.9860 - val_loss: 1.0141 Epoch 24/300 4/4 [==============================] - 0s 18ms/step - loss: 0.9843 - val_loss: 1.0124 Epoch 25/300 4/4 [==============================] - 0s 12ms/step - loss: 0.9825 - val_loss: 1.0111 Epoch 26/300 4/4 [==============================] - 0s 10ms/step - loss: 0.9808 - val_loss: 1.0093 Epoch 27/300 4/4 [==============================] - 0s 14ms/step - loss: 0.9791 - val_loss: 1.0081 Epoch 28/300 4/4 [==============================] - 0s 11ms/step - loss: 0.9774 - val_loss: 1.0060 Epoch 29/300 4/4 [==============================] - 0s 11ms/step - loss: 0.9755 - val_loss: 1.0045 Epoch 30/300 4/4 [==============================] - 0s 12ms/step - loss: 0.9738 - val_loss: 1.0031 Epoch 31/300 4/4 [==============================] - 0s 11ms/step - loss: 0.9723 - val_loss: 1.0016 Epoch 32/300 4/4 [==============================] - 0s 10ms/step - loss: 0.9708 - val_loss: 0.9997 Epoch 33/300 4/4 [==============================] - 0s 11ms/step - loss: 0.9688 - val_loss: 0.9980 Epoch 34/300 4/4 [==============================] - 0s 13ms/step - loss: 0.9671 - val_loss: 0.9963 Epoch 35/300 4/4 [==============================] - 0s 14ms/step - loss: 0.9653 - val_loss: 0.9945 Epoch 36/300 4/4 [==============================] - 0s 11ms/step - loss: 0.9635 - val_loss: 0.9930 Epoch 37/300 4/4 [==============================] - 0s 10ms/step - loss: 0.9619 - val_loss: 0.9916 Epoch 38/300 4/4 [==============================] - 0s 11ms/step - loss: 0.9604 - val_loss: 0.9897 Epoch 39/300 4/4 [==============================] - 0s 11ms/step - loss: 0.9583 - val_loss: 0.9881 Epoch 40/300 4/4 [==============================] - 0s 10ms/step - loss: 0.9566 - val_loss: 0.9867 Epoch 41/300 4/4 [==============================] - 0s 11ms/step - loss: 0.9551 - val_loss: 0.9846 Epoch 42/300 4/4 [==============================] - 0s 12ms/step - loss: 0.9532 - val_loss: 0.9829 Epoch 43/300 4/4 [==============================] - 0s 10ms/step - loss: 0.9514 - val_loss: 0.9811 Epoch 44/300 4/4 [==============================] - 0s 11ms/step - loss: 0.9496 - val_loss: 0.9795 Epoch 45/300 4/4 [==============================] - 0s 12ms/step - loss: 0.9479 - val_loss: 0.9777 Epoch 46/300 4/4 [==============================] - 0s 12ms/step - loss: 0.9461 - val_loss: 0.9761 Epoch 47/300 4/4 [==============================] - 0s 12ms/step - loss: 0.9445 - val_loss: 0.9743 Epoch 48/300 4/4 [==============================] - 0s 10ms/step - loss: 0.9425 - val_loss: 0.9726 Epoch 49/300 4/4 [==============================] - 0s 12ms/step - loss: 0.9407 - val_loss: 0.9712 Epoch 50/300 4/4 [==============================] - 0s 11ms/step - loss: 0.9392 - val_loss: 0.9691 Epoch 51/300 4/4 [==============================] - 0s 11ms/step - loss: 0.9371 - val_loss: 0.9673 Epoch 52/300 4/4 [==============================] - 0s 12ms/step - loss: 0.9353 - val_loss: 0.9656 Epoch 53/300 4/4 [==============================] - 0s 11ms/step - loss: 0.9336 - val_loss: 0.9638 Epoch 54/300 4/4 [==============================] - 0s 11ms/step - loss: 0.9316 - val_loss: 0.9619 Epoch 55/300 4/4 [==============================] - 0s 11ms/step - loss: 0.9299 - val_loss: 0.9606 Epoch 56/300 4/4 [==============================] - 0s 13ms/step - loss: 0.9282 - val_loss: 0.9585 Epoch 57/300 4/4 [==============================] - 0s 14ms/step - loss: 0.9262 - val_loss: 0.9567 Epoch 58/300 4/4 [==============================] - 0s 11ms/step - loss: 0.9243 - val_loss: 0.9548 Epoch 59/300 4/4 [==============================] - 0s 11ms/step - loss: 0.9225 - val_loss: 0.9530 Epoch 60/300 4/4 [==============================] - 0s 11ms/step - loss: 0.9207 - val_loss: 0.9513 Epoch 61/300 4/4 [==============================] - 0s 12ms/step - loss: 0.9187 - val_loss: 0.9493 Epoch 62/300 4/4 [==============================] - 0s 10ms/step - loss: 0.9168 - val_loss: 0.9475 Epoch 63/300 4/4 [==============================] - 0s 11ms/step - loss: 0.9149 - val_loss: 0.9459 Epoch 64/300 4/4 [==============================] - 0s 12ms/step - loss: 0.9131 - val_loss: 0.9441 Epoch 65/300 4/4 [==============================] - 0s 11ms/step - loss: 0.9110 - val_loss: 0.9419 Epoch 66/300 4/4 [==============================] - 0s 11ms/step - loss: 0.9091 - val_loss: 0.9401 Epoch 67/300 4/4 [==============================] - 0s 11ms/step - loss: 0.9071 - val_loss: 0.9383 Epoch 68/300 4/4 [==============================] - 0s 11ms/step - loss: 0.9052 - val_loss: 0.9365 Epoch 69/300 4/4 [==============================] - 0s 10ms/step - loss: 0.9033 - val_loss: 0.9343 Epoch 70/300 4/4 [==============================] - 0s 11ms/step - loss: 0.9013 - val_loss: 0.9325 Epoch 71/300 4/4 [==============================] - 0s 12ms/step - loss: 0.8993 - val_loss: 0.9303 Epoch 72/300 4/4 [==============================] - 0s 11ms/step - loss: 0.8972 - val_loss: 0.9284 Epoch 73/300 4/4 [==============================] - 0s 10ms/step - loss: 0.8952 - val_loss: 0.9263 Epoch 74/300 4/4 [==============================] - 0s 11ms/step - loss: 0.8931 - val_loss: 0.9243 Epoch 75/300 4/4 [==============================] - 0s 11ms/step - loss: 0.8910 - val_loss: 0.9221 Epoch 76/300 4/4 [==============================] - 0s 12ms/step - loss: 0.8890 - val_loss: 0.9203 Epoch 77/300 4/4 [==============================] - 0s 12ms/step - loss: 0.8869 - val_loss: 0.9181 Epoch 78/300 4/4 [==============================] - 0s 13ms/step - loss: 0.8847 - val_loss: 0.9160 Epoch 79/300 4/4 [==============================] - 0s 19ms/step - loss: 0.8827 - val_loss: 0.9139 Epoch 80/300 4/4 [==============================] - 0s 12ms/step - loss: 0.8804 - val_loss: 0.9116 Epoch 81/300 4/4 [==============================] - 0s 12ms/step - loss: 0.8783 - val_loss: 0.9095 Epoch 82/300 4/4 [==============================] - 0s 10ms/step - loss: 0.8761 - val_loss: 0.9072 Epoch 83/300 4/4 [==============================] - 0s 11ms/step - loss: 0.8738 - val_loss: 0.9049 Epoch 84/300 4/4 [==============================] - 0s 10ms/step - loss: 0.8716 - val_loss: 0.9030 Epoch 85/300 4/4 [==============================] - 0s 12ms/step - loss: 0.8694 - val_loss: 0.9003 Epoch 86/300 4/4 [==============================] - 0s 11ms/step - loss: 0.8670 - val_loss: 0.8982 Epoch 87/300 4/4 [==============================] - 0s 10ms/step - loss: 0.8647 - val_loss: 0.8959 Epoch 88/300 4/4 [==============================] - 0s 12ms/step - loss: 0.8623 - val_loss: 0.8934 Epoch 89/300 4/4 [==============================] - 0s 11ms/step - loss: 0.8599 - val_loss: 0.8911 Epoch 90/300 4/4 [==============================] - 0s 11ms/step - loss: 0.8577 - val_loss: 0.8886 Epoch 91/300 4/4 [==============================] - 0s 11ms/step - loss: 0.8552 - val_loss: 0.8862 Epoch 92/300 4/4 [==============================] - 0s 12ms/step - loss: 0.8527 - val_loss: 0.8837 Epoch 93/300 4/4 [==============================] - 0s 13ms/step - loss: 0.8502 - val_loss: 0.8812 Epoch 94/300 4/4 [==============================] - 0s 12ms/step - loss: 0.8478 - val_loss: 0.8788 Epoch 95/300 4/4 [==============================] - 0s 10ms/step - loss: 0.8453 - val_loss: 0.8762 Epoch 96/300 4/4 [==============================] - 0s 11ms/step - loss: 0.8428 - val_loss: 0.8736 Epoch 97/300 4/4 [==============================] - 0s 11ms/step - loss: 0.8401 - val_loss: 0.8711 Epoch 98/300 4/4 [==============================] - 0s 17ms/step - loss: 0.8376 - val_loss: 0.8686 Epoch 99/300 4/4 [==============================] - 0s 12ms/step - loss: 0.8349 - val_loss: 0.8660 Epoch 100/300 4/4 [==============================] - 0s 19ms/step - loss: 0.8322 - val_loss: 0.8634 Epoch 101/300 4/4 [==============================] - 0s 12ms/step - loss: 0.8295 - val_loss: 0.8605 Epoch 102/300 4/4 [==============================] - 0s 11ms/step - loss: 0.8269 - val_loss: 0.8580 Epoch 103/300 4/4 [==============================] - 0s 11ms/step - loss: 0.8240 - val_loss: 0.8552 Epoch 104/300 4/4 [==============================] - 0s 12ms/step - loss: 0.8213 - val_loss: 0.8524 Epoch 105/300 4/4 [==============================] - 0s 12ms/step - loss: 0.8185 - val_loss: 0.8495 Epoch 106/300 4/4 [==============================] - 0s 13ms/step - loss: 0.8157 - val_loss: 0.8466 Epoch 107/300 4/4 [==============================] - 0s 12ms/step - loss: 0.8128 - val_loss: 0.8437 Epoch 108/300 4/4 [==============================] - 0s 12ms/step - loss: 0.8099 - val_loss: 0.8407 Epoch 109/300 4/4 [==============================] - 0s 11ms/step - loss: 0.8070 - val_loss: 0.8378 Epoch 110/300 4/4 [==============================] - 0s 13ms/step - loss: 0.8042 - val_loss: 0.8351 Epoch 111/300 4/4 [==============================] - 0s 11ms/step - loss: 0.8012 - val_loss: 0.8319 Epoch 112/300 4/4 [==============================] - 0s 11ms/step - loss: 0.7983 - val_loss: 0.8290 Epoch 113/300 4/4 [==============================] - 0s 11ms/step - loss: 0.7953 - val_loss: 0.8261 Epoch 114/300 4/4 [==============================] - 0s 11ms/step - loss: 0.7923 - val_loss: 0.8230 Epoch 115/300 4/4 [==============================] - 0s 13ms/step - loss: 0.7893 - val_loss: 0.8199 Epoch 116/300 4/4 [==============================] - 0s 11ms/step - loss: 0.7862 - val_loss: 0.8168 Epoch 117/300 4/4 [==============================] - 0s 10ms/step - loss: 0.7833 - val_loss: 0.8138 Epoch 118/300 4/4 [==============================] - 0s 12ms/step - loss: 0.7803 - val_loss: 0.8106 Epoch 119/300 4/4 [==============================] - 0s 10ms/step - loss: 0.7772 - val_loss: 0.8073 Epoch 120/300 4/4 [==============================] - 0s 15ms/step - loss: 0.7742 - val_loss: 0.8039 Epoch 121/300 4/4 [==============================] - 0s 15ms/step - loss: 0.7711 - val_loss: 0.8006 Epoch 122/300 4/4 [==============================] - 0s 13ms/step - loss: 0.7680 - val_loss: 0.7974 Epoch 123/300 4/4 [==============================] - 0s 15ms/step - loss: 0.7650 - val_loss: 0.7938 Epoch 124/300 4/4 [==============================] - 0s 12ms/step - loss: 0.7620 - val_loss: 0.7904 Epoch 125/300 4/4 [==============================] - 0s 12ms/step - loss: 0.7588 - val_loss: 0.7869 Epoch 126/300 4/4 [==============================] - 0s 11ms/step - loss: 0.7558 - val_loss: 0.7837 Epoch 127/300 4/4 [==============================] - 0s 11ms/step - loss: 0.7527 - val_loss: 0.7801 Epoch 128/300 4/4 [==============================] - 0s 12ms/step - loss: 0.7498 - val_loss: 0.7768 Epoch 129/300 4/4 [==============================] - 0s 11ms/step - loss: 0.7468 - val_loss: 0.7738 Epoch 130/300 4/4 [==============================] - 0s 11ms/step - loss: 0.7439 - val_loss: 0.7708 Epoch 131/300 4/4 [==============================] - 0s 11ms/step - loss: 0.7412 - val_loss: 0.7678 Epoch 132/300 4/4 [==============================] - 0s 12ms/step - loss: 0.7384 - val_loss: 0.7645 Epoch 133/300 4/4 [==============================] - 0s 13ms/step - loss: 0.7357 - val_loss: 0.7614 Epoch 134/300 4/4 [==============================] - 0s 13ms/step - loss: 0.7330 - val_loss: 0.7585 Epoch 135/300 4/4 [==============================] - 0s 12ms/step - loss: 0.7304 - val_loss: 0.7554 Epoch 136/300 4/4 [==============================] - 0s 12ms/step - loss: 0.7277 - val_loss: 0.7526 Epoch 137/300 4/4 [==============================] - 0s 10ms/step - loss: 0.7253 - val_loss: 0.7500 Epoch 138/300 4/4 [==============================] - 0s 11ms/step - loss: 0.7229 - val_loss: 0.7470 Epoch 139/300 4/4 [==============================] - 0s 11ms/step - loss: 0.7204 - val_loss: 0.7444 Epoch 140/300 4/4 [==============================] - 0s 11ms/step - loss: 0.7181 - val_loss: 0.7418 Epoch 141/300 4/4 [==============================] - 0s 11ms/step - loss: 0.7157 - val_loss: 0.7392 Epoch 142/300 4/4 [==============================] - 0s 14ms/step - loss: 0.7134 - val_loss: 0.7365 Epoch 143/300 4/4 [==============================] - 0s 12ms/step - loss: 0.7112 - val_loss: 0.7342 Epoch 144/300 4/4 [==============================] - 0s 11ms/step - loss: 0.7088 - val_loss: 0.7319 Epoch 145/300 4/4 [==============================] - 0s 12ms/step - loss: 0.7067 - val_loss: 0.7297 Epoch 146/300 4/4 [==============================] - 0s 12ms/step - loss: 0.7046 - val_loss: 0.7275 Epoch 147/300 4/4 [==============================] - 0s 11ms/step - loss: 0.7026 - val_loss: 0.7258 Epoch 148/300 4/4 [==============================] - 0s 10ms/step - loss: 0.7007 - val_loss: 0.7239 Epoch 149/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6988 - val_loss: 0.7220 Epoch 150/300 4/4 [==============================] - 0s 13ms/step - loss: 0.6970 - val_loss: 0.7204 Epoch 151/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6953 - val_loss: 0.7187 Epoch 152/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6936 - val_loss: 0.7170 Epoch 153/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6920 - val_loss: 0.7153 Epoch 154/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6905 - val_loss: 0.7137 Epoch 155/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6891 - val_loss: 0.7121 Epoch 156/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6876 - val_loss: 0.7109 Epoch 157/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6863 - val_loss: 0.7091 Epoch 158/300 4/4 [==============================] - 0s 13ms/step - loss: 0.6852 - val_loss: 0.7073 Epoch 159/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6839 - val_loss: 0.7061 Epoch 160/300 4/4 [==============================] - 0s 17ms/step - loss: 0.6829 - val_loss: 0.7049 Epoch 161/300 4/4 [==============================] - 0s 14ms/step - loss: 0.6819 - val_loss: 0.7036 Epoch 162/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6811 - val_loss: 0.7024 Epoch 163/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6803 - val_loss: 0.7014 Epoch 164/300 4/4 [==============================] - 0s 13ms/step - loss: 0.6795 - val_loss: 0.7004 Epoch 165/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6790 - val_loss: 0.7000 Epoch 166/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6785 - val_loss: 0.6990 Epoch 167/300 4/4 [==============================] - 0s 17ms/step - loss: 0.6779 - val_loss: 0.6983 Epoch 168/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6774 - val_loss: 0.6969 Epoch 169/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6768 - val_loss: 0.6961 Epoch 170/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6763 - val_loss: 0.6952 Epoch 171/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6760 - val_loss: 0.6947 Epoch 172/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6754 - val_loss: 0.6937 Epoch 173/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6750 - val_loss: 0.6932 Epoch 174/300 4/4 [==============================] - 0s 13ms/step - loss: 0.6745 - val_loss: 0.6928 Epoch 175/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6741 - val_loss: 0.6921 Epoch 176/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6736 - val_loss: 0.6915 Epoch 177/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6733 - val_loss: 0.6910 Epoch 178/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6727 - val_loss: 0.6908 Epoch 179/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6724 - val_loss: 0.6904 Epoch 180/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6720 - val_loss: 0.6900 Epoch 181/300 4/4 [==============================] - 0s 13ms/step - loss: 0.6717 - val_loss: 0.6898 Epoch 182/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6714 - val_loss: 0.6896 Epoch 183/300 4/4 [==============================] - 0s 13ms/step - loss: 0.6710 - val_loss: 0.6894 Epoch 184/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6707 - val_loss: 0.6890 Epoch 185/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6701 - val_loss: 0.6884 Epoch 186/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6699 - val_loss: 0.6882 Epoch 187/300 4/4 [==============================] - 0s 13ms/step - loss: 0.6695 - val_loss: 0.6877 Epoch 188/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6692 - val_loss: 0.6873 Epoch 189/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6688 - val_loss: 0.6867 Epoch 190/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6684 - val_loss: 0.6864 Epoch 191/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6683 - val_loss: 0.6862 Epoch 192/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6679 - val_loss: 0.6857 Epoch 193/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6674 - val_loss: 0.6854 Epoch 194/300 4/4 [==============================] - 0s 14ms/step - loss: 0.6670 - val_loss: 0.6852 Epoch 195/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6671 - val_loss: 0.6848 Epoch 196/300 4/4 [==============================] - 0s 13ms/step - loss: 0.6664 - val_loss: 0.6845 Epoch 197/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6663 - val_loss: 0.6843 Epoch 198/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6660 - val_loss: 0.6838 Epoch 199/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6654 - val_loss: 0.6833 Epoch 200/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6652 - val_loss: 0.6835 Epoch 201/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6654 - val_loss: 0.6827 Epoch 202/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6649 - val_loss: 0.6829 Epoch 203/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6647 - val_loss: 0.6827 Epoch 204/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6643 - val_loss: 0.6823 Epoch 205/300 4/4 [==============================] - 0s 15ms/step - loss: 0.6640 - val_loss: 0.6819 Epoch 206/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6640 - val_loss: 0.6816 Epoch 207/300 4/4 [==============================] - 0s 15ms/step - loss: 0.6636 - val_loss: 0.6819 Epoch 208/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6636 - val_loss: 0.6816 Epoch 209/300 4/4 [==============================] - 0s 14ms/step - loss: 0.6634 - val_loss: 0.6809 Epoch 210/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6630 - val_loss: 0.6806 Epoch 211/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6628 - val_loss: 0.6810 Epoch 212/300 4/4 [==============================] - 0s 13ms/step - loss: 0.6628 - val_loss: 0.6806 Epoch 213/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6624 - val_loss: 0.6807 Epoch 214/300 4/4 [==============================] - 0s 13ms/step - loss: 0.6621 - val_loss: 0.6804 Epoch 215/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6621 - val_loss: 0.6800 Epoch 216/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6618 - val_loss: 0.6800 Epoch 217/300 4/4 [==============================] - 0s 15ms/step - loss: 0.6616 - val_loss: 0.6799 Epoch 218/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6615 - val_loss: 0.6792 Epoch 219/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6613 - val_loss: 0.6788 Epoch 220/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6612 - val_loss: 0.6785 Epoch 221/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6609 - val_loss: 0.6786 Epoch 222/300 4/4 [==============================] - 0s 13ms/step - loss: 0.6609 - val_loss: 0.6786 Epoch 223/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6606 - val_loss: 0.6784 Epoch 224/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6605 - val_loss: 0.6782 Epoch 225/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6604 - val_loss: 0.6782 Epoch 226/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6602 - val_loss: 0.6781 Epoch 227/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6599 - val_loss: 0.6780 Epoch 228/300 4/4 [==============================] - 0s 16ms/step - loss: 0.6599 - val_loss: 0.6782 Epoch 229/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6596 - val_loss: 0.6779 Epoch 230/300 4/4 [==============================] - 0s 17ms/step - loss: 0.6596 - val_loss: 0.6777 Epoch 231/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6593 - val_loss: 0.6773 Epoch 232/300 4/4 [==============================] - 0s 18ms/step - loss: 0.6593 - val_loss: 0.6774 Epoch 233/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6590 - val_loss: 0.6772 Epoch 234/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6588 - val_loss: 0.6772 Epoch 235/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6587 - val_loss: 0.6769 Epoch 236/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6585 - val_loss: 0.6768 Epoch 237/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6584 - val_loss: 0.6762 Epoch 238/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6582 - val_loss: 0.6763 Epoch 239/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6581 - val_loss: 0.6762 Epoch 240/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6580 - val_loss: 0.6766 Epoch 241/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6580 - val_loss: 0.6761 Epoch 242/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6577 - val_loss: 0.6762 Epoch 243/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6577 - val_loss: 0.6758 Epoch 244/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6575 - val_loss: 0.6759 Epoch 245/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6574 - val_loss: 0.6758 Epoch 246/300 4/4 [==============================] - 0s 13ms/step - loss: 0.6573 - val_loss: 0.6757 Epoch 247/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6571 - val_loss: 0.6760 Epoch 248/300 4/4 [==============================] - 0s 13ms/step - loss: 0.6571 - val_loss: 0.6755 Epoch 249/300 4/4 [==============================] - 0s 13ms/step - loss: 0.6569 - val_loss: 0.6754 Epoch 250/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6567 - val_loss: 0.6755 Epoch 251/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6566 - val_loss: 0.6757 Epoch 252/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6565 - val_loss: 0.6760 Epoch 253/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6564 - val_loss: 0.6755 Epoch 254/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6563 - val_loss: 0.6756 Epoch 255/300 4/4 [==============================] - 0s 14ms/step - loss: 0.6562 - val_loss: 0.6754 Epoch 256/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6561 - val_loss: 0.6754 Epoch 257/300 4/4 [==============================] - 0s 14ms/step - loss: 0.6560 - val_loss: 0.6753 Epoch 258/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6559 - val_loss: 0.6756 Epoch 259/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6557 - val_loss: 0.6751 Epoch 260/300 4/4 [==============================] - 0s 13ms/step - loss: 0.6558 - val_loss: 0.6748 Epoch 261/300 4/4 [==============================] - 0s 13ms/step - loss: 0.6555 - val_loss: 0.6747 Epoch 262/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6555 - val_loss: 0.6746 Epoch 263/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6554 - val_loss: 0.6750 Epoch 264/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6554 - val_loss: 0.6753 Epoch 265/300 4/4 [==============================] - 0s 13ms/step - loss: 0.6553 - val_loss: 0.6750 Epoch 266/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6551 - val_loss: 0.6754 Epoch 267/300 4/4 [==============================] - 0s 14ms/step - loss: 0.6550 - val_loss: 0.6746 Epoch 268/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6548 - val_loss: 0.6747 Epoch 269/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6549 - val_loss: 0.6749 Epoch 270/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6548 - val_loss: 0.6748 Epoch 271/300 4/4 [==============================] - 0s 13ms/step - loss: 0.6547 - val_loss: 0.6747 Epoch 272/300 4/4 [==============================] - 0s 13ms/step - loss: 0.6546 - val_loss: 0.6745 Epoch 273/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6545 - val_loss: 0.6743 Epoch 274/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6545 - val_loss: 0.6746 Epoch 275/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6543 - val_loss: 0.6750 Epoch 276/300 4/4 [==============================] - 0s 18ms/step - loss: 0.6543 - val_loss: 0.6746 Epoch 277/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6542 - val_loss: 0.6742 Epoch 278/300 4/4 [==============================] - 0s 13ms/step - loss: 0.6540 - val_loss: 0.6742 Epoch 279/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6540 - val_loss: 0.6742 Epoch 280/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6538 - val_loss: 0.6739 Epoch 281/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6538 - val_loss: 0.6735 Epoch 282/300 4/4 [==============================] - 0s 13ms/step - loss: 0.6536 - val_loss: 0.6736 Epoch 283/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6535 - val_loss: 0.6735 Epoch 284/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6534 - val_loss: 0.6739 Epoch 285/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6534 - val_loss: 0.6737 Epoch 286/300 4/4 [==============================] - 0s 10ms/step - loss: 0.6531 - val_loss: 0.6737 Epoch 287/300 4/4 [==============================] - 0s 13ms/step - loss: 0.6532 - val_loss: 0.6738 Epoch 288/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6531 - val_loss: 0.6733 Epoch 289/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6528 - val_loss: 0.6736 Epoch 290/300 4/4 [==============================] - 0s 19ms/step - loss: 0.6529 - val_loss: 0.6738 Epoch 291/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6528 - val_loss: 0.6732 Epoch 292/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6527 - val_loss: 0.6730 Epoch 293/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6525 - val_loss: 0.6733 Epoch 294/300 4/4 [==============================] - 0s 13ms/step - loss: 0.6525 - val_loss: 0.6732 Epoch 295/300 4/4 [==============================] - 0s 13ms/step - loss: 0.6524 - val_loss: 0.6733 Epoch 296/300 4/4 [==============================] - 0s 13ms/step - loss: 0.6523 - val_loss: 0.6732 Epoch 297/300 4/4 [==============================] - 0s 13ms/step - loss: 0.6523 - val_loss: 0.6729 Epoch 298/300 4/4 [==============================] - 0s 12ms/step - loss: 0.6521 - val_loss: 0.6731 Epoch 299/300 4/4 [==============================] - 0s 11ms/step - loss: 0.6521 - val_loss: 0.6729 Epoch 300/300 4/4 [==============================] - 0s 13ms/step - loss: 0.6520 - val_loss: 0.6728 32/32 [==============================] - 0s 2ms/step 32/32 [==============================] - 0s 1ms/step 32/32 [==============================] - 0s 1ms/step 32/32 [==============================] - 0s 2ms/step
# plot history
fig, axes = plt.subplots(ncols=4, figsize=(25, 6))
axes[0].plot(history1.history['loss'], label='Training loss')
axes[0].plot(history1.history['val_loss'], label='Validation loss')
axes[1].plot(history2.history['loss'], label='Training loss')
axes[1].plot(history2.history['val_loss'], label='Validation loss')
axes[2].plot(history3.history['loss'], label='Training loss')
axes[2].plot(history3.history['val_loss'], label='Validation loss')
axes[3].plot(history4.history['loss'], label='Training loss')
axes[3].plot(history4.history['val_loss'], label='Validation loss')
axes[0].legend()
axes[0].set_title('Class 1')
axes[0].set(
ylabel='Loss',
xlabel='Epoch'
)
axes[1].set_title('Class 2')
axes[1].legend()
axes[1].set(
ylabel='Loss',
xlabel='Epoch'
)
axes[2].set_title('Class 3')
axes[2].legend()
axes[2].set(
ylabel='Loss',
xlabel='Epoch'
)
axes[3].set_title('Class 4')
axes[3].legend()
axes[3].set(
ylabel='Loss',
xlabel='Epoch'
)
plt.show()
# plot the data
plt.scatter(encoded_x1[:,0], encoded_x1[:,1], label='Class 1')
plt.scatter(encoded_x2[:,0], encoded_x2[:,1], label='Class 2')
plt.scatter(encoded_x3[:,0], encoded_x3[:,1], label='Class 3')
plt.scatter(encoded_x4[:,0], encoded_x4[:,1], label='Class 4')
plt.legend()
plt.show()
Pesquise sobre redes neurais recorrentes LSTM. Apresente neste estudo aplicações das LSTM deep learning. Seguem abaixo sugestões de aplicações.
As redes neurais LSTM (Long Short-Term Memory) são um tipo específico de redes neurais recorrentes que receberam muita atenção recentemente na comunidade de aprendizado de máquina. De forma geral, as redes LSTM possuem a característica de realimentação, o que permite um efeito de memória de curto e longo prazo. Com isso, a saída é modulada pelo estado dessas células, tornando-se uma propriedade muito importante quando precisamos que a previsão da rede neural dependa do contexto histórico das entradas, e não apenas da última entrada.
Exemplo de uma aplicação
Uma das aplicações das redes neurais LSTM é a predição de séries temporais, isto é, tentar prever o próximo valor a partir de valores anteriores. Assim, escolheu-se um problema onde, dado um ano e um mês, a tarefa é prever o número de passageiros de companhias aéreas internacionais em unidades de 1.000. Os dados variam de janeiro de 1949 a dezembro de 1960, ou 12 anos, com 144 observações.
Ou seja, dado o número de passageiros (em unidades de milhares) neste mês, qual é o número de passageiros no próximo mês?
Você pode escrever uma função simples para converter a única coluna de dados em um conjunto de dados de duas colunas: a primeira coluna contendo a contagem de passageiros deste mês (t) e a segunda coluna contendo a contagem de passageiros do próximo mês (t+1) a ser prevista.
# LSTM for international airline passengers problem with regression framing
import numpy as np
import matplotlib.pyplot as plt
from pandas import read_csv
import math
import tensorflow as tf
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Dense
from tensorflow.keras.layers import LSTM
from sklearn.preprocessing import MinMaxScaler
from sklearn.metrics import mean_squared_error
# convert an array of values into a dataset matrix
def create_dataset(dataset, look_back=1):
dataX, dataY = [], []
for i in range(len(dataset)-look_back-1):
a = dataset[i:(i+look_back), 0]
dataX.append(a)
dataY.append(dataset[i + look_back, 0])
return np.array(dataX), np.array(dataY)
# fix random seed for reproducibility
tf.random.set_seed(7)
# load the dataset
dataframe = read_csv('airline-passengers.csv', usecols=[1], engine='python')
dataset = dataframe.values
dataset = dataset.astype('float32')
# normalize the dataset
scaler = MinMaxScaler(feature_range=(0, 1))
dataset = scaler.fit_transform(dataset)
# split into train and test sets
train_size = int(len(dataset) * 0.67)
test_size = len(dataset) - train_size
train, test = dataset[0:train_size,:], dataset[train_size:len(dataset),:]
print("Train shape: ", train.shape)
print("Test shape: ", test.shape)
Train shape: (96, 1) Test shape: (48, 1)
# reshape into X=t and Y=t+1
look_back = 1
trainX, trainY = create_dataset(train, look_back)
testX, testY = create_dataset(test, look_back)
# reshape input to be [samples, time steps, features]
trainX = np.reshape(trainX, (trainX.shape[0], 1, trainX.shape[1]))
testX = np.reshape(testX, (testX.shape[0], 1, testX.shape[1]))
# create and fit the LSTM network
model = Sequential()
model.add(LSTM(4, input_shape=(1, look_back)))
model.add(Dense(1))
model.compile(loss='mean_squared_error', optimizer='adam')
model.fit(trainX, trainY, epochs=100, batch_size=1, verbose=2)
Epoch 1/100 94/94 - 2s - loss: 0.0431 - 2s/epoch - 24ms/step Epoch 2/100 94/94 - 0s - loss: 0.0229 - 224ms/epoch - 2ms/step Epoch 3/100 94/94 - 0s - loss: 0.0163 - 212ms/epoch - 2ms/step Epoch 4/100 94/94 - 0s - loss: 0.0148 - 232ms/epoch - 2ms/step Epoch 5/100 94/94 - 0s - loss: 0.0140 - 217ms/epoch - 2ms/step Epoch 6/100 94/94 - 0s - loss: 0.0130 - 217ms/epoch - 2ms/step Epoch 7/100 94/94 - 0s - loss: 0.0121 - 217ms/epoch - 2ms/step Epoch 8/100 94/94 - 0s - loss: 0.0112 - 226ms/epoch - 2ms/step Epoch 9/100 94/94 - 0s - loss: 0.0104 - 211ms/epoch - 2ms/step Epoch 10/100 94/94 - 0s - loss: 0.0094 - 220ms/epoch - 2ms/step Epoch 11/100 94/94 - 0s - loss: 0.0084 - 211ms/epoch - 2ms/step Epoch 12/100 94/94 - 0s - loss: 0.0074 - 219ms/epoch - 2ms/step Epoch 13/100 94/94 - 0s - loss: 0.0066 - 221ms/epoch - 2ms/step Epoch 14/100 94/94 - 0s - loss: 0.0057 - 217ms/epoch - 2ms/step Epoch 15/100 94/94 - 0s - loss: 0.0050 - 214ms/epoch - 2ms/step Epoch 16/100 94/94 - 0s - loss: 0.0043 - 214ms/epoch - 2ms/step Epoch 17/100 94/94 - 0s - loss: 0.0037 - 220ms/epoch - 2ms/step Epoch 18/100 94/94 - 0s - loss: 0.0032 - 216ms/epoch - 2ms/step Epoch 19/100 94/94 - 0s - loss: 0.0029 - 210ms/epoch - 2ms/step Epoch 20/100 94/94 - 0s - loss: 0.0026 - 220ms/epoch - 2ms/step Epoch 21/100 94/94 - 0s - loss: 0.0024 - 212ms/epoch - 2ms/step Epoch 22/100 94/94 - 0s - loss: 0.0022 - 226ms/epoch - 2ms/step Epoch 23/100 94/94 - 0s - loss: 0.0021 - 208ms/epoch - 2ms/step Epoch 24/100 94/94 - 0s - loss: 0.0020 - 218ms/epoch - 2ms/step Epoch 25/100 94/94 - 0s - loss: 0.0021 - 217ms/epoch - 2ms/step Epoch 26/100 94/94 - 0s - loss: 0.0020 - 208ms/epoch - 2ms/step Epoch 27/100 94/94 - 0s - loss: 0.0020 - 220ms/epoch - 2ms/step Epoch 28/100 94/94 - 0s - loss: 0.0020 - 215ms/epoch - 2ms/step Epoch 29/100 94/94 - 0s - loss: 0.0021 - 219ms/epoch - 2ms/step Epoch 30/100 94/94 - 0s - loss: 0.0020 - 210ms/epoch - 2ms/step Epoch 31/100 94/94 - 0s - loss: 0.0020 - 222ms/epoch - 2ms/step Epoch 32/100 94/94 - 0s - loss: 0.0021 - 221ms/epoch - 2ms/step Epoch 33/100 94/94 - 0s - loss: 0.0021 - 213ms/epoch - 2ms/step Epoch 34/100 94/94 - 0s - loss: 0.0020 - 214ms/epoch - 2ms/step Epoch 35/100 94/94 - 0s - loss: 0.0021 - 213ms/epoch - 2ms/step Epoch 36/100 94/94 - 0s - loss: 0.0020 - 224ms/epoch - 2ms/step Epoch 37/100 94/94 - 0s - loss: 0.0020 - 207ms/epoch - 2ms/step Epoch 38/100 94/94 - 0s - loss: 0.0020 - 213ms/epoch - 2ms/step Epoch 39/100 94/94 - 0s - loss: 0.0020 - 219ms/epoch - 2ms/step Epoch 40/100 94/94 - 0s - loss: 0.0020 - 223ms/epoch - 2ms/step Epoch 41/100 94/94 - 0s - loss: 0.0020 - 216ms/epoch - 2ms/step Epoch 42/100 94/94 - 0s - loss: 0.0020 - 216ms/epoch - 2ms/step Epoch 43/100 94/94 - 0s - loss: 0.0020 - 212ms/epoch - 2ms/step Epoch 44/100 94/94 - 0s - loss: 0.0020 - 210ms/epoch - 2ms/step Epoch 45/100 94/94 - 0s - loss: 0.0021 - 216ms/epoch - 2ms/step Epoch 46/100 94/94 - 0s - loss: 0.0019 - 229ms/epoch - 2ms/step Epoch 47/100 94/94 - 0s - loss: 0.0020 - 216ms/epoch - 2ms/step Epoch 48/100 94/94 - 0s - loss: 0.0021 - 216ms/epoch - 2ms/step Epoch 49/100 94/94 - 0s - loss: 0.0020 - 211ms/epoch - 2ms/step Epoch 50/100 94/94 - 0s - loss: 0.0020 - 219ms/epoch - 2ms/step Epoch 51/100 94/94 - 0s - loss: 0.0020 - 213ms/epoch - 2ms/step Epoch 52/100 94/94 - 0s - loss: 0.0020 - 209ms/epoch - 2ms/step Epoch 53/100 94/94 - 0s - loss: 0.0020 - 211ms/epoch - 2ms/step Epoch 54/100 94/94 - 0s - loss: 0.0020 - 219ms/epoch - 2ms/step Epoch 55/100 94/94 - 0s - loss: 0.0020 - 210ms/epoch - 2ms/step Epoch 56/100 94/94 - 0s - loss: 0.0020 - 215ms/epoch - 2ms/step Epoch 57/100 94/94 - 0s - loss: 0.0020 - 213ms/epoch - 2ms/step Epoch 58/100 94/94 - 0s - loss: 0.0020 - 214ms/epoch - 2ms/step Epoch 59/100 94/94 - 0s - loss: 0.0020 - 219ms/epoch - 2ms/step Epoch 60/100 94/94 - 0s - loss: 0.0020 - 212ms/epoch - 2ms/step Epoch 61/100 94/94 - 0s - loss: 0.0020 - 217ms/epoch - 2ms/step Epoch 62/100 94/94 - 0s - loss: 0.0021 - 213ms/epoch - 2ms/step Epoch 63/100 94/94 - 0s - loss: 0.0020 - 211ms/epoch - 2ms/step Epoch 64/100 94/94 - 0s - loss: 0.0020 - 221ms/epoch - 2ms/step Epoch 65/100 94/94 - 0s - loss: 0.0020 - 224ms/epoch - 2ms/step Epoch 66/100 94/94 - 0s - loss: 0.0020 - 217ms/epoch - 2ms/step Epoch 67/100 94/94 - 0s - loss: 0.0021 - 209ms/epoch - 2ms/step Epoch 68/100 94/94 - 0s - loss: 0.0020 - 219ms/epoch - 2ms/step Epoch 69/100 94/94 - 0s - loss: 0.0020 - 209ms/epoch - 2ms/step Epoch 70/100 94/94 - 0s - loss: 0.0020 - 214ms/epoch - 2ms/step Epoch 71/100 94/94 - 0s - loss: 0.0020 - 214ms/epoch - 2ms/step Epoch 72/100 94/94 - 0s - loss: 0.0020 - 216ms/epoch - 2ms/step Epoch 73/100 94/94 - 0s - loss: 0.0020 - 222ms/epoch - 2ms/step Epoch 74/100 94/94 - 0s - loss: 0.0021 - 229ms/epoch - 2ms/step Epoch 75/100 94/94 - 0s - loss: 0.0020 - 227ms/epoch - 2ms/step Epoch 76/100 94/94 - 0s - loss: 0.0020 - 211ms/epoch - 2ms/step Epoch 77/100 94/94 - 0s - loss: 0.0019 - 218ms/epoch - 2ms/step Epoch 78/100 94/94 - 0s - loss: 0.0020 - 216ms/epoch - 2ms/step Epoch 79/100 94/94 - 0s - loss: 0.0020 - 209ms/epoch - 2ms/step Epoch 80/100 94/94 - 0s - loss: 0.0020 - 211ms/epoch - 2ms/step Epoch 81/100 94/94 - 0s - loss: 0.0019 - 216ms/epoch - 2ms/step Epoch 82/100 94/94 - 0s - loss: 0.0021 - 220ms/epoch - 2ms/step Epoch 83/100 94/94 - 0s - loss: 0.0020 - 211ms/epoch - 2ms/step Epoch 84/100 94/94 - 0s - loss: 0.0019 - 217ms/epoch - 2ms/step Epoch 85/100 94/94 - 0s - loss: 0.0020 - 214ms/epoch - 2ms/step Epoch 86/100 94/94 - 0s - loss: 0.0020 - 223ms/epoch - 2ms/step Epoch 87/100 94/94 - 0s - loss: 0.0021 - 235ms/epoch - 2ms/step Epoch 88/100 94/94 - 0s - loss: 0.0020 - 231ms/epoch - 2ms/step Epoch 89/100 94/94 - 0s - loss: 0.0020 - 213ms/epoch - 2ms/step Epoch 90/100 94/94 - 0s - loss: 0.0020 - 217ms/epoch - 2ms/step Epoch 91/100 94/94 - 0s - loss: 0.0020 - 228ms/epoch - 2ms/step Epoch 92/100 94/94 - 0s - loss: 0.0020 - 210ms/epoch - 2ms/step Epoch 93/100 94/94 - 0s - loss: 0.0020 - 226ms/epoch - 2ms/step Epoch 94/100 94/94 - 0s - loss: 0.0020 - 209ms/epoch - 2ms/step Epoch 95/100 94/94 - 0s - loss: 0.0021 - 216ms/epoch - 2ms/step Epoch 96/100 94/94 - 0s - loss: 0.0020 - 219ms/epoch - 2ms/step Epoch 97/100 94/94 - 0s - loss: 0.0020 - 214ms/epoch - 2ms/step Epoch 98/100 94/94 - 0s - loss: 0.0020 - 215ms/epoch - 2ms/step Epoch 99/100 94/94 - 0s - loss: 0.0020 - 213ms/epoch - 2ms/step Epoch 100/100 94/94 - 0s - loss: 0.0020 - 221ms/epoch - 2ms/step
<keras.callbacks.History at 0x7f663c5adb10>
# make predictions
trainPredict = model.predict(trainX)
testPredict = model.predict(testX)
3/3 [==============================] - 0s 4ms/step 2/2 [==============================] - 0s 5ms/step
# invert predictions
trainPredict = scaler.inverse_transform(trainPredict)
trainY = scaler.inverse_transform([trainY])
testPredict = scaler.inverse_transform(testPredict)
testY = scaler.inverse_transform([testY])
# calculate root mean squared error
trainScore = np.sqrt(mean_squared_error(trainY[0], trainPredict[:,0]))
print('Train Score: %.2f RMSE' % (trainScore))
testScore = np.sqrt(mean_squared_error(testY[0], testPredict[:,0]))
print('Test Score: %.2f RMSE' % (testScore))
Train Score: 22.68 RMSE Test Score: 49.34 RMSE
# shift train predictions for plotting
trainPredictPlot = np.empty_like(dataset)
trainPredictPlot[:, :] = np.nan
trainPredictPlot[look_back:len(trainPredict)+look_back, :] = trainPredict
# shift test predictions for plotting
testPredictPlot = np.empty_like(dataset)
testPredictPlot[:, :] = np.nan
testPredictPlot[len(trainPredict)+(look_back*2)+1:len(dataset)-1, :] = testPredict
# plot baseline and predictions
plt.plot(scaler.inverse_transform(dataset))
plt.plot(trainPredictPlot)
plt.plot(testPredictPlot)
plt.show()
Apresente um estudo sobre transferência de conhecimento (transfer learning) no contexto de deep learning.
De acordo com Brownlee (2019), Transfer Learning é um método de aprendizagem de máquina no qual permite que um modelo treinado para uma determinada tarefa seja utilizado como um ponto de partida de um modelo para outro problema com características similares.
Esta técnica é bastante utilizada em tarefas de visão computacional, processamento de linguagem natural, dados os vastos recursos computacionais (principalmente as GPUs) e tempo necessários para desenvolver modelos de rede neural sobre esses problemas, além das emissões de CO2 causadas pelo gasto de energia. Além disso, é possível encontrar modelos treinados por grandes empresas por muitas horas e com uma grande infraestrutura disponíveis de forma gratuita na internet. Isso permite obter excelentes resultados com um tempo de treinamento muito menor do que seria treinar um novo modelo do zero.
É possível utilizar esta técnica a partir dos três passos a seguir:
Selecione Modelo pré-treinado: Escolha um modelo de origem pré-treinado dentre os modelos disponíveis. Muitas instituições de pesquisa lançam modelos em conjuntos de dados grandes e desafiadores que podem ser incluídos no conjunto de modelos candidatos a serem escolhidos.
Modelo de Reutilização: Utiliza o modelo pré-treinado como ponto de partida para um modelo na segunda tarefa de interesse. Isso pode envolver o uso de todo ou de partes do modelo, dependendo da técnica de modelagem utilizada.
Modelo de sintonia: Opcionalmente, o modelo pode precisar ser adaptado ou refinado nos dados do par de entrada-saída disponíveis para a tarefa de interesse. Assim, você pode adicionar mais camadas ao final da rede para adptar ao seu problema.
Exemplos de modelos pré-treinados na área de visão computacional:
Exemplos de modelos pré-treinados na área de processamento de linguagem natural: